Dissertation submitted to the Combined Faculties for the Natural Sciences and for Mathematics of the Ruperto-Carola University of Heidelberg, Germany for the degree of Doctor of Natural Sciences
presented by
Diplom-Physicist: Lothar Keck born in: Maulbronn
Oral examination: 21st November, 2001
Climate significance
of stable isotope records from Alpine ice cores
Referees: Prof. Dr. Ulrich Platt
Prof. Dr. Konrad Mauersberger
Climate significance of Alpine ice core stable isotope records
Stable water isotope records (δ18O and δD) of three down to bedrock ice cores from Monte Rosa massif (Swiss Alps) were evaluated in view of potentially recorded temperature signals. On-line measurements of δD by a novel water reduction method yield an accuracy of 0.7 ‰. Upstream effects in δ18O ice core records, assessed from recent surface composition and modelled backward trajectories, may decrease the 20th century trend by a factor of three. The comparison of δ18O records with 240 years of instrumental growing season temperatures indicates that major decadal and centennial temperature changes are concurrently recorded with a ∆δ18O/∆T relation of 1.7 ‰/°C. Decadal means of growing season isotope temperatures indicate that recent levels are the highest, though not unprecedented within the last millennium. The unparalleled respective minimum occurred around 1340 A.D. with decadal means of more than 1°C below the recent level. The three cores exhibit an isotopically depleted basal layer. Contrary to the hitherto assumption of their Pleistocene origin, a new hypothesis is introduced that explains these signals as a consequence of the rapid deformation of glacier ice near bedrock.
Klimatische Aussagekraft der stabilen Wasserisotopomere in Alpinen Eisbohrkernen
Tiefenprofile der stabilen Wasserisotopomere (δ18O und δD) dreier tiefer Eisbohrkerne vom Monte Rosa Massiv (Schweizer Alpen) wurden im Hinblick auf möglicherweise aufgezeichnete Temperatursignale ausgewertet. Durch eine neuartige Methode der Wasserreduktion sind On-line δD Messungen mit einer Genauigkeit von 0.7 ‰ möglich. Upstream-Effekte in den δ18O Profilen, abgeschätzt aus der gegenwärtigen Zusammensetzung der Gletscheroberfläche und modellierten Rückwärtstrajektorien, können die Trends des 20. Jahrhunderts um einen Faktor drei vermindern. Der Vergleich von δ18O Profilen mit instrumentellen Sommertemperaturen über 240 Jahre zeigt, daß Temperaturtrends über eine Zeitskala von Jahrzehnten bis Jahrhunderten übereinstimmend aufgezeichnet sind, wobei die Empfindlichkeit ∆δ18O/∆T etwa bei 1.7 ‰/°C liegt. Die derzeitigen mehrjährigen Mittel der Isotopentemperatur sind mit die wärmsten innerhalb des vergangenen Jahrtausends. Das beispiellose Temperaturminimum um 1340 A.D. weist um mehr als 1°C niedrigere mehrjährige Mittel der Isotopentemperatur auf. Alle drei Kerne zeigen eine isotopisch leichte basale Schicht. Im Gegensatz zu der bisherigen Annahme ihres eiszeitlichen Ursprungs wird eine neue Theorie aufgestellt, welche diese Signale als Folge der schnelleren Eisdeformation in Felsbettnähe erklärt. Contents
LIST OF ABBREVIATIONS...... 9
1. INTRODUCTION...... 10
2. STABLE WATER ISOTOPES IN PRECIPITATION...... 17
2.1 FRACTIONATION PROCESSES ...... 17 2.2 FORMATION OF PRECIPITATION ...... 17 2.2.1 Equilibrium condensation...... 17 2.2.2 Kinetic effects ...... 18 2.2.3 Sub-cloud effects...... 20 2.3 GENERATION OF ATMOSPHERIC WATER VAPOUR...... 20 2.4 VARIABILITY IN ALPINE PRECIPITATION...... 21 2.4.1 Seasonal variations and altitude effect...... 21 2.4.2. Long term ∆δ18O/∆T -relations...... 24
3. METHODS ...... 26
3.1 CORE DRILLING AND PROCESSING ...... 26 3.2 HYDROGEN ISOTOPES...... 27 3.2.1 Principle of the H/Device ...... 28 3.2.2 Adaptation of the method to the MAT 230C mass spectrometer...... 28 3.2.3 Effects of apparative parameters...... 29 3.2.4 Precision...... 32 3.2.5 Memory effect ...... 32 3.2.6 Calibration ...... 32 3.2.7 Standard drift...... 33 3.2.8 Overall Reproducibility ...... 33 3.2.9 Summary ...... 34 3.3 OXYGEN ISOTOPES ...... 34 3.4 STABLE ISOTOPE INTERCOMPARISIONS...... 34
4. GLACIOLOGICAL FEATURES...... 36
4.1 STEADY ICE FLOW ...... 36 4.1.1 Ice deformation...... 36 4.1.2 Flow models for CG...... 38 4.2 MASS BALANCE...... 39 4.2.1 Background...... 39 4.2.2 Mass balance measurement...... 40 4.2.3 Results and discussion ...... 43
5. DATING...... 45
5.1 STATE OF THE DATING...... 45 KCS...... 45 KCH...... 48 CC...... 48 5.2 PROGRESS DURING THIS WORK...... 48 KCS...... 49 KCH...... 50 CC...... 51 5.3 SUMMARY...... 54
6. SAMPLING TRAIT...... 56
6.1 RECORDED SEASON ...... 56 6.1.1 Melt events...... 56 6.1.2 Wind erosion...... 58 6.1.3 Seasonal variations in precipitation amount ...... 59 6.1.4 Application on ice core records...... 61 6.2 RECORDED WEATHER TYPES ...... 64 6.2.1 Precipitation amount ...... 64 6.2.2 Fresh snow samples...... 65 6.3 SUMMARY...... 68
7. STABLE ISOTOPE LONG TERM RECORDS ...... 69
7.1 DATED SEQUENCES ...... 69 7.2 SYSTEMATIC UPSTREAM EFFECTS...... 72 7.2.2 Backward trajectories...... 72 7.2.2 Shallow cores...... 74 7.2.3 Result ...... 76 7.2.4 Summary and discussion...... 77 7.3 COMPARISON WITH INSTRUMENTAL RECORDS ...... 78 7.3.1 Site specific temperature series ...... 78 7.3.2 Qualitative comparison ...... 79 7.2.4 Sensitivity of the stable isotope thermometer...... 82 7.4 A MILLENNIUM OF ISOTOPE TEMPERATURES ...... 83 7.5 COMPARISON TO OTHER MILLENNIAL RECORDS ...... 85 7.5.1 Oscillations of Glaciers ...... 85 7.5.2 Tree-rings ...... 85 7.6 SENSITIVITY ...... 87 7.6.1 The restriction on the growing season...... 87 7.6.2 The high elevation ...... 87 7.6.3 Climate induced changes in the ice core seasonal composition ...... 91 7.6.4 Climate induced changes in precipitation types ...... 91 7.6.5 Summary ...... 91 7.7 REMARKS ON THE D-EXCESS...... 92 7.8 SUMMARY...... 92
8. BASAL LAYER ISOTOPE SIGNALS ...... 94
8.1 RELATION TO OTHER SITES...... 94 8.2 PLEISTOCENE ICE...... 96 8.2.1 Trace gases in air bubbles...... 96 8.2.2 Other potential age constraints ...... 98 8.3 SOURCE REGION ANOMALY ...... 98 8.3.1 Mass flow rates...... 100 8.3.2 Isotopic composition...... 100 8.4 MIGRATION OF ISOTOPES...... 102 8.4.1 Heuristic preliminary remark ...... 102 8.4.2 Stratigraphy and density...... 103 8.4.3 Ice structure and fabric ...... 106 8.4.4 The liquid phase in polycrystalline cold ice ...... 106 8.4.5 Suggested process...... 107 8.4.6 Predicted effect on isotope profiles...... 107 8.4.7 Relation to Ion content...... 112 8.5 SUMMARY AND OUTLOOK ...... 113
REFERENCES...... 115
ANNEX A: CORE META DATA...... I
ANNEX B: UPDATED VOLCANO CHRONOLOGY...... II
ANNEX C: RADIONUCLIDE INVENTORIES...... V
C.1 BACKGROUND ...... V C.1.1 The natural radioisotopes 10Be and 210Pb ...... V C.1.2 The artificial radioisotopes 137Cs and Tritium ...... VII C.2 METHODS ...... VIII C.3 RESULTS...... IX
ANNEX D: GLACIER SIMULATOR EXPERIMENTS ...... XI
D.1 PRODUCTION OF HOMOGENEOUS POLYCRISTALLINE ICE ...... XI D.2 DESIGN AND OPERATION OF THE GLACIER SIMULATOR...... XI D.3 RESULTS AND DISCUSSION...... XIII List of abbreviations
Drill sites CDD Col du Dome, Mont Blanc massif CDL Col du Lys, Monte Rosa massif CG Colle Gnifetti, MonteRosa massif
Ice cores CC Down to bedrock core at Colle Gnifetti KCH do. KCS do.
Institutes DISAT Dipartimento di Scienze dell’ Ambiente e del Territorio, Milano DISGAM Dipartimento di Scienze Geologiche, Ambientali e Marine ETHZ Laboratory of Hydraulics, Hydrology and Glaciology, Zurich IUP Institut für Umweltphysik, Heidelberg LGGE Laboratoire de Glaciologie et Geophysique de’l Environnement, St. Martin d’ Heres LMCE Laboratoire de Modélisation du Climat et de l'Environnement, Saclay ZAMG Central Institute for Meteorology and Geodynamics, Vienna
Analyses CFA Continuous flow analyses ECM Electric conductivity measurement IC Ion chromatography MS Mass spectrometer
Other ALPCLIM EU project: “Environmental and Climate Records from High Elevation Alpine Glaciers” IAEA International Atomic Energy Agency SSA Singular spectrum analyses 10 1. Introduction
1. Introduction
Environment and climate have always strongly affected the fate of man. Nowadays in turn human activities may cause climate change because the radiation balance of the earth is affected by anthropogenic trace gases and aerosols which were intensely produced since the beginning of the industrial era (Shine et al. 1990). Thus there is need to asses any potential influence of human activities on climate against the past natural variability. Assuming model derived natural fluctuations the global warming in the second half of the 20th century is most probably already attributable to the human influence (Tett 1999). However, global instrumental observations cover at best the last 150 years with respect to air temperature and only a few decades with respect to atmospheric trace gases and aerosols. Therefore temporal and spatial patterns of the natural variability in climate and geochemical cycles, as well as their interactions, can only poorly be estimated from instrumental records. To extent the time range of atmospheric records various types of natural climate archives, as sediments, trees and glaciers, can be exploited. The most comprehensive records are obtained from ice cores of “cold” glaciers. Such glaciers are characterised by ice temperatures well below the pressure melting point and thus the records are not blurred by meltwater 18 percolation. The use of the stable water isotope (H2 O or HDO) content of the ice matrix as 18 temperature proxy is well established (Dansgaard 1973). Concurrent measurements of H2 O and HDO providing the secondary quantity deuterium-excess may can give insight into the water vapour source region of the stored precipitation (Barlow et al. 1993). Moreover the atmospheric aerosol load is recorded in the impurity content of the ice and atmospheric gases are trapped in air bubbles. Deep ice cores from polar ice sheets yielded fundamental new insight in the climate system. The most outstanding results are currently the reconstruction of 420.000 years of isotope temperature and atmospheric trace gase composition from the Vostok ice core, Antarctica (Petit et al. 1999) and high resolution records of one full glacial-interglacial cycle from the GRIP (Dansgaard et al. 1993) and GISP2 (Stuiver and Grootes 2000) ice cores, central Greenland. However, while these two Greenland records show close agreement in isotope pattern for the Holocene and most part of the Wisconsin glaciation, significant differences in the lower core sections spanning the Eemiam interglacial and the penultimate glaciation remain until now unexplained (Grootes et al. 1993). The success of polar ice core studies motivates the investigation of ice cores from alpine1 glaciers to access the potentially conserved most valuable mid- and low latitude climate records. Specifically the European Alps are most intriguing for ice core studied for two reasons:
1 Note that “alpine” is used for high mountains in general, whereas “Alpine” refers to the European Alps 11
• In contrast to all other ice core sites, a dense network of instrumental meteorological records in the European Alps and adjacent regions over more than 250 years is available. This offers the unique possibility of a mutual validation of ice core climate proxies and instrumental long term records. • Due to the comparably short life time of atmospheric particles the anthropogenic change in aerosol load and its possible impact on climate is concentrated on industrialised areas. Thus ice core records from cold Alpine glaciers are particular due to their location within major antropogenic source areas. However, Alpine cold glaciers exist solely in the high elevation summit regions, since only there the mean annual air temperature (MAAT) is low enough. As a consequence, Alpine drill sites have several peculiarities as pointed out by Wagenbach (1992): • Due to their exposed location on small scale glaciers in the summit regions, they are subject to strong wind erosion of snow. As the extend of wind erosion depends on meteorological conditions and surface geometry, it is spatially and seasonally variable. Consequently, the finally conserved fraction of annual precipitation may reflect a strongly seasonally weighted atmospheric signal. • The high altitude aerosol concentration is controlled by the seasonally variable intensity of vertical mixing. Therefore the seasonal amplitudes of aerosol concentrations are always large compared to their long term trends. This applies also to the stable isotopes in mid- latitude precipitation. As a result, long term trends of stable water isotopes or aerosol related records derived from alpine ice cores are very sensitive to any changes in the conserved fraction of annual precipitation. • Due to the local sources of trace substances and water vapour, the seasonally variable vertical mixing intensity is expected to cause seasonal variations in the spatial representativity of Alpine ice core records. • The conserved fraction of annual precipitation generally undergoes systematic spatial changes at the glacier surface. Due to the inflow of ice from upstream the drill site, temporal trends from alpine ice cores are superimposed by the spatial variations along the surface. Such upstream effects are to be taken into account even if the core is drilled at a saddle point (Wagner 1996). • Cold alpine glaciers in the summit regions are of irregular topography, with the horizontal extension being comparable to glacier thickness. Particularly in the proximity to bedrock this causes a complex ice flow pattern and a strongly non-linear age-depth relation. Due to the annual layer thinning seasonal signals in aerosol concentrations can not be resolved in the lower core sections. Thus the dating of the major parts of the time range remains a challenge. The EU-project ALPCLIM (Environmental and Climate Records from High Elevation Alpine Glaciers) was devoted to the exploitation of Alpine ice cores for climate related record. It was launched in 1998 and coordinated by the Institut für Umweltphysik (IUP), Universität 12 1. Introduction
Heidelberg. The sampling approach of ALPCLIM is a multi-core study at three Alpine drill sites indicated in Figure 1.1: • Colle Gnifetti (CG, located in the Monte Rosa massif) • Col du Lys (CDL, as well Monte Rosa massif) • Col du Dome (CDD, Mont Blanc massif)
Figure.1.1: Dots: Location of the main ALPCLIM drill sites Colle Gnifetti and Col du Lys at Monte Rosa (45° 56’ N, 7° 52’ E) and Col du Dome at Mont Blanc (45° 50’ N, 6° 50’ E).
The main concern of this thesis are the ice cores from CG, a small firn saddle situated at 4450 m asl. As illustrated in Figure 1.2, CG is located in an extremely exposed location. Therefore strong wind erosion causes an extraordinary low annual accumulation of snow. As a result, CG is the unique Alpine key site to recover long term ice core records. Accordingly CG is glacialogically well characterised (Alean et al. 1984) and was subject to ice core studies and related work since the 70ties (Oeschger et al. 1977). However, the specific problems of Alpine drill sites apply in particular for CG. 13
Figure 1.2: View of Colle Gnifetti (indicated by the arrow) from north. The three summits of Monte Rosa range are (from left to right) Signalkuppe, Zumsteinspitze and Nordend.
The shape of the saddle, being situated between Zumsteinspitze and Signalkuppe, is shown in Figure 1.3. The annual accumulation varies systematically from 0.1 m w.e.2 at the shadily north-west slope of Signalkuppe to more than 1.2 m w.e. at the sunny south slope of Zumsteinspitze, where the formation of ice layers effectively protects the snow from wind erosion (Alean et al. 1983). The wind erosion affects preferably the isotopically light dry winter snow (Wagenbach 1989). Thereby the systematic variation of annual accumulation is accompanied by a spatial variations of the mean stable isotope content, as illustrated in Figure 1.4. This spatial variations are large compared to Holocene stable isotope long term trends. Thus, due to the potential upstream effects e.g. an increase of stable isotopes in the second half of the 20th century which had been observed in the “Chemie Core” (CC, drilled in 1982 at the position shown in Figure 1.3) can not unambiguously be attributed to a temperature signal. In fact the observed increase is much stronger than expected from the simultaneous warming trend in Alpine air temperatures.
2 m w.e. (meter water equivalent) is the thickness of a layer of snow or firn when compressed to the density of water. The unit is commonly used to account for the compaction of snow and firn. 14 1. Introduction
Zumsteinspitze
Ice cliff
4500
¡ ¢
¡ ¡
¡ £
00 44 I Grenzgletscher I I I I I I I I I I I I I I 0 I 50 Signalkuppe I 4 I I I I I I II I I I I I I I I I I I I I I
I
¥ ¤ ¤ ¦ ¤ ¤ § ¤ I I I I I I I I I
Figure 1.3: Surface topography of CG. The deep cores KCH and KCS (drilled in 1995) and CC (drilled in 1982) are aligned approximately along a surface flow line.
The annual accumulation of snow is also strongly affected by surface topography. The topography is changed by wind action and potentially by glacier flow. This causes both quasi- random spatial variations of the annual net snow accumulation and systematic temporal variations in annual snow accumulation at fixed location. Such variations in annual accumulation in ice cores from CG have been reported by Armbruster (2000). A systematic increase of annual accumulation with air temperature was suggested by Schotterer et al. (1978). The presence of non steady state conditions at Colle Gnifetti is also supported by photos showing visible changes of the surface topography at the ice cliff and at the rock outcrops in the slope of Signalkuppe during the last decades (Wagenbach pers. com.). On the other hand, a digital comparison of a photograph taken in 1883 with a recent one taken from the same position shows no discernible changes for the bulk surface topography (Wagner 1996). In any case, potential temporal changes in surface topography are a matter of concern as the annual snow accumulation and as shown in Figure 1.4 also the stable isotope content would be affected. 15
-10 1.0 δ18O annual accumulation ] . e . w
18 m δ [ O long term trends ] n o i ‰ t [
a -15 0.5 l O u 8 1 m δ u c c a
l a u n
Increasing fraction of winter snow n a
-20 0.0 -200 0 200 400 Position [m]
Figure 1.4: Spatial variations of annual accumulation and surface isotopic composition between Zumsteinspitze (at position - 350 m), saddle point (at 0 m) and Signalkuppe (at + 400 m)
The useful time scale in ice cores from CG covers at least several centuries (Schäfer 1995). However, the basal layer section of cold Alpine glaciers, which is expected to contain the major part of the conserved time range, has not yet been dated. Flow models suggest an age of a few thousand years for the basal ice (Wagner 1996) whereas stable isotope anomalies hint at potential remnants of Pleistocene ice (Wagenbach 1993). This would be in agreement with the time horizon of alpine ice core records from the Tibetan Plateau (Thompson et al. 1997) and the Andes (Thompson et al. 1998), from where climate and environmental records back into the last glacial stage are reported. To exploit the unique climate records of CG, two down to bedrock cores (KCH and KCS) were drilled in 1995 by IUP in collaboration with Physikalisches Institut, Universität Bern (PIUB). As displayed in Figure 1.3, these cores are approximately on a surface flow line with the already existing and completely analysed core CC from 1982. Within ALPCLIM, one core was recovered at Colle de Lys (CDL), situated at 4150 m asl in the southern part of the Monte Rosa massiv and only 1 km from CG by the Dipartimento di Scienze dell’Ambiente e del Territorio (DISAT), Milano. As the annual snow accumulation is relatively well preserved at CDL, the record is of seasonal resolution though covering only the last three decades. Finally, an ice core from Col du Dôme (CDD), situated at 4250 m asl in the northern part of the Mont Blanc summit range, was available. The core thence had been drilled and mainly analyses by the Laboratoire de Glaciologie et Géophysique de l’Environnement (LGGE). The accumulation rate is comparable to CDL and thus also here ice core records in sub-seasonal resolution could be obtained over 80 years, approximately. 16 1. Introduction
All cores were comprehensively analysed on stable isotopes and major ions. To link the records, the inventories of the radionuclides 10Be, 210Pb, 3H and 137Cs, each with a relatively well known spatial and temporal pattern of source and transport, were measured for each core. Moreover, within ALPCLIM Alpine temperature series back to 1760 A.D. had been collected, homogenised and gridded (Böhm et al. 2000) by the Zentralanstalt für Meteorologie and Geodynamik (ZAMG), Vienna. Furthermore for each drill site a detailed recent climatological characterisation was established by ZAMG as well. Additionally, englacial temperature profiles as a potential additional temperature proxy in the Alpine region were evaluated by Laboratory of Hydraulics, Hydrology and Glaciology of ETH Zürich (ETHZ). To assess climate records from the CG stable isotope long term records, the interpretation effort of this thesis focuses on the interconnected basic problems of Alpine ice core records: • To assess their seasonal representativity of the ice records. This is mandatory both to compare of stable isotope records with instrumental time series and to estimate the spatial representativity of inherent temperature signals. • To evaluate the influence of systematic upstream effects on stable isotope long term trends. This required intense field work for the drilling of supplementary shallow cores and the measurement of surface mass balance. • To validate the ice core climate proxies by independent information from instrumental temperature series. In particular a direct calibration of the stable isotope thermometer is envisaged. The latter is a crucial point for the interpretation since the sensitivity of the stable isotope thermometer strongly depends on the location. This is concluded both by direct observations (Rozanski et al. 1997) and model studies (Cole et al. 1999). However, this requires a precise dating and an appropriate processing of the various data sets. The validation of stable isotope trends from CG is attempted by comparing them with the respective results of the other ALPCLIM drill sites. On the base of these investigations the basic prospects of temperature proxies from Alpine ice core records can eventually be assessed. A part of this work arose of the reflection on the basal ice at CG. While the light stable isotope values suggest pleistocene ice, the visual stratigraphy suggested an alternative explanation by the migration of water between ice grains into shear planes. This hypothesis on the origin of the main isotope signal in alpine ice cores is presented and discussed. 17
2. Stable water isotopes in precipitation
2.1 Fractionation processes
Besides the ordinary water molecule H2O of mass 18 amu, also small concentrations of the 18 main heavy isotopomeres HDO and H2 O participate in the hydrological cycle. The 18 variations of HDO and H2 O concentrations are controlled by the fractionation during evaporation and condensation. With R1 and R2 being the isotope ratio in phase 1 and phase 2 18 (R = [H2 O]/[H2O] or R= [HDO]/[H2O], respectively) the fractionation factor is defined as: R α = 1 R2 A measure for the degree of fractionation is ε, defined as ε = α - 1. Because the absolute isotope ratios R are hard to measure and in order to have a standardised scale, it is valuable to describe the isotopic composition of natural water bodies on the δ-scale versus V-SMOW. This δ-value in permille is defined as: R − R δ = sample V −SMOW ⋅ V −SMOW 1000 RV −SMOW
V-SMOW is an artificial standard water with an isotopic composition being approximately equal to the Standard Mean Ocean Water (SMOW). The fractionation at phase transitions is based on two physical processes: 1. Equilibrium fractionation: Since the vapour pressure of the heavier components is slightly lower, vapour is generally depleted in heavy isotope content compared to a liquid or solid phase in equilibrium. 2. Kinetic effects: Heavier isotopes exhibit a lower molecular diffusion coefficient. Thus they have to overcome a larger transfer resitance when passing through a laminar-viscose 18 18 layer. As the equilibrium fractionation is much lower for H2 O, the variations in δ O concentrations are proportionally more affected by kinetic effects than for δD. Kinetic effects, which occur both at the evaporation and condensation, are the main cause that the two isotopes do not behave completely analogous in the natural water cycle.
2.2 Formation of precipitation
2.2.1 Equilibrium condensation During the cooling of an air parcel, the heterogeneous condensation of its water vapour content starts and is maintained at a very low supersaturation of a few permille. Thus kinetic 18 2. Stable water isotopes in precipitation
effects at the formation of liquid cloud drops are normally negligible. When condensation starts, the newly formed condensate is enriched in heavy isotopes while the remaining vapour is accordingly depleted. If the cooling proceeds and the condensed portion is removed from the air parcel, the remaining water vapour undergoes a successive depletion in heavy isotopes content and thus also the continuously newly formed condensate. The degree of the depletion is controlled by the remaining fraction of the initial water vapour content and thus by the temperature difference between initial dew point and actual temperature of the air parcel. This implies all the so-called isotope effects as there are seasonal-, latitude-, altitude-, inland-, the paleo-climatic effect (Dansgaard 1973) and the amount effect. The so called “simple Raleigh condensation model“ assumes that the generation of liquid phase takes place in equilibrium and that the condensate is immediately removed from the cloud without any isotope exchange with the vapour. This simple Raleigh model is applicable only on clouds with low liquid water content. The model yields (Dansgaard 1964): • The ∆δ18O/∆T -relation, in other words the sensitivity of the isotope thermometer with respect to the ambient temperature T, increases at low temperatures. Typical values are between 0.24 ‰ /°C (solid-liquid condensation and moist adiabatic cooling at 20°C) and 0.82 ‰/°C (vapour-solid condensation and isobaric cooling at 0°C). • The slope ∆δD/∆δ18O is affected both by the instantaneous composition of water vapour and the temperature dependent fractionation factors. In the range of moderate atmospheric temperatures the effects counterbalance such that the slope is almost constantly 8 both for isobaric and adiabatic cooling ((2) → (3) in Figure 2.1). Thus the impact of non- equilibrium effects is reflected in the deuterium excess (D-excess or d), being defined as
d = δD − 8 ⋅δ 18O
2.2.2 Kinetic effects The simple equilibrium Raleigh model with vapour-solid fractionation factors predicts a pronounced increase of the D-excess at polar temperatures ((3) → (4a) in Figure 2.1). Contrary to that, only a gradual increase has been observed in polar snow ((3) → (4) in Figure 2.1). Moreover, the model predicted ∆δ18O/∆T –relation strongly exceeds the observations. Model predictions and field data can be reconciled by taking into account kinetic effects at the formation of snow (Jouzel and Merlivat 1984). 19
δD [‰] MWL (d=10) d=0 -30 -20 -10 (2) 18 (SMOW) δ O [‰]
liquid or solid (1) vapour (3a) (3) -100
-200
(4a) (4)
Figure 2.1: Variations of δ18O and δD in at evaporation and condensation. The D- excess (d) is the vertical distance to the straight line δD = 8⋅ δ18O. MWL: meteoric water line. (SMOW)→(1): Evaporation from the sea surface. (1)→(2): condensation of water vapour at dew point (2)→(3): condensation at moderate temperatures (3)→(3a): evaporation of raindrops (3)→(4a): condensation at low temperatures as predicted by equilibrium condensation (not observed) (3)→(4) condensation at low temperatures taking into account kinetic effects
More generally, at temperatures between the activation of ice nuclei and the homogeneous nucleation of water droplets, solid and liquid phase generally coexist in mixed cloud systems. Since the saturation vapour pressure with respect to ice is lower than with respect to water, the ice crystals grow on the expense of the liquid droplets (Rogers 1989). During this so called Bergeron-Findesein process kinetic effects occur both at evaporation of droplets and at condensation on ice crystals. While for the δ-values this leads only to minor changes compared with the predictions of the vapour-solid kinetic model, it remedies instabilities for the predicted D-excess with respect to cloud vapour pressure history (Ciais and Jouzel 1994). The δ-values and the D-excess in precipitation also depend on the temperature at which clouds switch from supercooled drops to ice particles. Therefore, also the microparticle load of an air parcel acting as freezing nuclei affect the isotopic composition of precipitation (Fisher 1991). However, contrary to the formation of polar precipitation in stratiform clouds it has been suggested that the stable isotopes in precipitation from deep convective clouds are almost unaffected by the Bergeron-Findesein process (Federer et al. 1982). The process of supercooled liquid water droplets in clouds being captured by ice particles is called riming. Case studies indicate that depending on the vertical air motion in clouds riming 20 2. Stable water isotopes in precipitation
may take place preferably in the lower and warmer (Demoz et al. 1991) or higher and cooler parts of the cloud (Fujiyoshi et al. 1986). Correspondingly the stable isotope values of rimed particles may over- or underestimate the mean cloud condensation temperature.
2.2.3 Sub-cloud effects Raindrops may undergo a significant shift in their isotopic composition between cloud base and ground due to evaporation and isotopic exchange with ambient water vapour. As indicated in Figure 2.1 (3)→(3a) sub-cloud evaporation increases the δ-values and due to the kinetic effects it lowers the D-excess of precipitation (Ehhalt et al. 1963). In contrast to raindrops the effect of evaporation on the isotopic composition of snow flakes is negligible due to the low molecular diffusion rates in ice.
2.3 Generation of atmospheric water vapour
Both equilibrium fractionation and kinetic effects occur at the evaporation of water vapour from sea surface or lakes (Ehhalt and Knott 1964). The isotopic composition of the evaporating water vapour net flux depends on the local meteorological conditions temperature, humidity and wind speed. Moreover, due to the atmosphere-sea isotope exchange, it is also affected by the isotope ratio of the already existing vapour in the marine boundary layer (Craig and Gordon 1965). Assuming a global closed system, it can be shown that the D-excess in atmospheric water vapour increases with decreasing relative humidity h and increasing sea surface temperature SST (Merlivat und Jouzel 1979). For a relative humidity of 70 – 80 % and a SST of 20 – 25 °C, average values for the subtropical evaporation belts, the model predicts a D-excess of ≈ 10 ‰ ((SMOW) → (1) in Figure 2.1). In fact this value resembles the present day global average of the D-excess in precipitation. As the subsequent formation of precipitation does frequently hardly affects the D-excess, most meteoric water bodies are close to the meteoric water line (MWL, see Figure 2.1):
δD = 8⋅δ18O + 10
In agreement with the model a high D-excess occurs in areas of evaporation at low humidity, such as the eastern Mediterranean (Gat and Carmi 1970). However, the theoretically presupposed closure equation does not hold on a local scale (Jouzel and Koster 1996). At continental sites the re-evaporation of precipitation from plants, soil and surface waters particularly in summer may be a significant contribution to the local water vapour (Koster et al. 1993, Jakob and Sonntag 1991). Whereas the evaporation from land surface and in particular plant transpiration acts mainly without fractionation (Knott 1964, Zimmermann 1967), kinetic effects at the evaporation of meteoric surface waters can cause an enhanced D- excess in precipitation at the luv side of large lakes (Gat et al. 1994). 21
2.4 Variability in Alpine precipitation
The actual isotopic composition of water vapour in an air parcel reflects its history of condensation and admixture of fresh water vapour. Moreover, the relation between condensation temperature and surface temperature depends on the weather type. Accordingly the variable synoptic weather types cause a poor correlation of stable isotopes and local surface temperature for single precipitation events. However, monthly or annual means of stable isotopes in precipitation at a given point represent the average from a variety of precipitation events.. As this tends to average out the effects from different air mass histories, a positive correlation both for monthly (i.e. seasonal) and annual means in δ18O and local temperature is found for most mid- and high latitude sites (Rozanski et al. 1982). These temporal means are controlled by the spatial and temporal pattern of air mass trajectories and the precipitation-evapotranspiration history
2.4.1 Seasonal variations and altitude effect δ-values show distinct seasonal variations in the Alpine region. The seasonal cycle of δ 18O from the four Swiss IAEA precipitation series (IAEA 1981) is displayed in Figure 2.2. Like the seasonal variations in temperature the monthly means exhibit a maximum in July and August. However, compared with the seasonal variations in temperature, the variations in δ18O exhibits a steeper decrease in autumn and a more gradual increase in spring
-5 Bern (572 m asl.) Meiringen (632 m asl.) Guttannen (1055 m asl.) Grimsel (1950 m asl.) average
] -10 ‰ [
O 8 1 δ
-15
1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 2.2: Seasonal variations of δ18O in Alpine precipitation from Swiss IAEA stations based on the monthly means from 1971-1992. Except Bern all stations are located in the eastern Bernese Oberland. 22 2. Stable water isotopes in precipitation
A potential decrease of the amplitude in seasonal variations in δ18O due to the evaporation of raindrops preferably at the low altitude in summer was discussed by Siegentaler and Qeschger (1980). However, with the time series until 1992 according to Figure 2.2, neither the seasonal amplitude nor the shape of the cycle exhibits a systematic change with altitude. Moreover, since the percentage of solid precipitation increases with altitude the strength of a potential evaporation effect is expected to be of minor importance there. On the other hand, the ∆δ/∆T – relation predicted by a simple Raleigh model increases at low temperatures. Therefore the seasonal amplitude is expected to increase with altitude. The seasonal variations in δ 18O from the IAEA stations (altitude below 1850 m asl.) correspond to temporal seasonal ∆δ/∆T – relations of 0.33 – 0.54 ‰/°C (Rozanski et al. 1992). However, a value of 0.68 ‰/°C is reported from precipitation samples at Jungfraujoch (3580 m asl) (Schotterer et al. 1997). This suggests that in fact the seasonal amplitude increases with altitude. Based on long term means an overall altitude effect of 0.2 ‰/100 m is reported for the Bernese Alps. This relation holds up to an altitude of 4000 m asl. (Stichler and Schotterer 2000). Taking this value of 0.2 ‰/100 m the mean seasonal variation from the IAEA data can be shifted to the altitude of CG (4500 m asl.). However, also a spatial shift in δ18O values is expected to occur between in the Bernese Alps and the Monte Rosa massif as they are located on different major orographic divides of the Alps. Moreover, the regional altitude effect could be different in the Monte Rosa massif since the altitude effect for individual precipitation events is stronger at the luv side of the Alps (Dray et al. 1997). Therefore the extrapolated seasonal variations are compared to site specific mean δ18O means.
Grenzgletscher IAEA extrapolated
-15 ] ‰ [
O 8 1 δ -20
-25 1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 2.3: Seasonal variations of δ18O for the Monte Rosa massif at 4200 m asl. Solid: estimated from IAEA low level data and an altitude effects of 0.2 ‰/100 m. Dotted: monthly means from an ice core at Grenzgletscher. 23
25 fresh snow samples were collected in summer (middle of June to middle of September) 1998, 1999 and 2000 at Colle Gnifetti. They exhibit a mean δ18O of –13.6 ‰. This agrees well with the high altitude summer level estimated from the IAEA data shifted with the altitude effect of 0.2 ‰/100 m ( –14.1 ‰). The seasonal variations from IAEA data corrected on the altitude effects can also be compared to monthly δ18O means derived from an ice core at Grenzgletscher. This site is only 1 km from Colle Gnifetti at an altitude of 4200 m asl. (Eichler et al. 2001). With an annual accumulation of 2.7 m w.e. this core may display the seasonal variations completely. However, these ice core derived monthly means suffer from the uncertain dissection in months. Figure 2.3 shows that the seasonal variations from IAEA data shifted to the altitude of 4200 m asl. agree fairly well with the seasonal variations from the ice core with the latter suggesting a slightly lower winter minimum.. As being supported from site specific data, the mean seasonal variations from the IAEA data set extrapolated to an altitude of 4500 m asl. will be used as a best guess for the seasonal variations of δ18O in precipitation at Colle Gnifetti. The seasonal variations of the D-excess in high altitudes are hard to estimate. Figure 2.4 shows the seasonal variation of the for low level stations. The maximum in autumn may be explained by the evaporation into dry polar air masses over the Atlantic ocean. The summer level (mid of June to mid of September) of 9.4 ‰ is well below the respective value from fresh snow samples of 13 ‰. The difference may partly be explained by the evaporation of raindrops in summer. However, it may also suggest a potential influence of local uplift of air masses on the D-excess. An increasing D-excess during solid-liquid condensation at low temperatures is expected. This would increase the D-excess mainly in winter.
13 Bern Meiringen
12
] 11 ‰ [
s s e c
x 10 e - D 9
8
1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 2.4 Seasonal variation in D-excess at Alpine low level IAEA stations 24 2. Stable water isotopes in precipitation
2.4.2. Long term ∆δ18O/∆T -relations The correlation of annual means in δ18O and surface temperature varies substantially even on a regional scale (Rozanski et al 1992). Rozanski et al. (1992) estimated long term ∆δ18O/∆T - relations from four Swiss IAEA stations to be 0.45 – 1.1 ‰/°C. Long term ∆δ18O/∆T - relations based on ice core records are discussed in detail in chapter 7. The long term ∆δ18O/∆T –relation at a certain location could be strongly affected by continental water recycling as a significant portion of the variability in annual means of δ18O is caused by it (Koster et al. 1993). Specifically for summer precipitation in Alpine ice cores a strong effect of continental water recycling is expected since: • The contribution of continental re-evaporation is more important in summer, as being reflected by the weak inland effect in Europe during this season (Knott 1964). • The degree of local water recycling is expected to be relatively large in the Alps due to the persistence of local wind fields decoupled from the synoptic atmospheric circulation. Specifically at clear sky conditions and flat pressure distributions, thermic wind fields develop (Barry 1981). This contribution of local evaporation causes a distinct diurnal variations in the humidity field of Alpine valleys (Hennemuth and Neureither 1986). Evapo-transpiration of plants as the major part of continental evaporation acts mainly without fractionation. Therefore the shift of δ18O means due to continental water recycling is expected to be positive if the δ18O in groundwater is above the δ18O level of advected water vapour. Typical δ18O values of Alpine groundwater and central European water vapour are displayed in Table 2.1.
Table 2.1: δ18O in groundwater and seasonal range of central European water vapour. The mean isotopic composition of the Rhone river is assumed to approximate the central Alpine groundwater.
water body δ18O [‰]
groundwater at northern margin of the Alps (Sonntag et al. 1979) -10 Rhone river (Schotterer et al. 2000) -14.5 Heidelberg water vapour summer level (Agemar et al. 2001) -15.3 Heidelberg water vapour winter level (dto) -20.9
The magnitude of the inland effect in precipitation for Europe suggests that the values for Heidelberg reflects the isotopic composition of water vapour advected to the Alps from west. As δ18O in groundwater at the margin of the Alps is well above the values for advected vapour, evaporation from there is expected to increases the δ18O levels. However, due to the altitude effect the δ18O level in groundwater of Alpine valleys is much lower. As it is 25
comparable to the summer level of the advected vapour, the effect of evaporation from alpine valleys on the δ18O levels in summer is expected to be rather small. 26 3. Methods
3. Methods
3.1 Core drilling and processing
The two deep cores KCH and KCS were electromechanically drilled in October 1995. Core diameter is 3″ (≈ 7.5 cm). High resolution profiles of ECM (electric conductivity measurement) were established as a proxy for acid content. Density profiles were measured by gamma absorption at the Alfred-Wegener-Institut (AWI) in Bremerhaven. After that the cores were cut lengthwise using a band saw as shown in Figure 3.1. Additional shallow cores were drilled between 1998 and 2000 using a Sarbach 4″ (≈ 10 cm) drill driven by petrol engine. For these cores only the routine analyses of stable isotopes, ion content and partly liquid conductivity were performed.
5 RI 2 0 IC SI SI
2 Surface 5 Archiv CFA for ECM 2 5
16 43 16
Figure 3.1: Sample cut pattern for the deep cores KCH and KCS. SI: stable isotopes, IC: ion chromatography and visual stratigraphy, RI: radioisotopes, CFA: continuos flow analyses.
An overview of the depth resolution for each sample type is given in Table 3.1. For the stable isotope samples the outer surface of the core was scraped off to eliminate potential effects of sublimation. The decontamination of the samples for chemical analyses was performed using an electric plane as described by Fischer (1997). Ion chromatography (IC) for Anions (Cl,
NO3, SO4) and Cations (Na, NH4, K, Mg, Ca), accompanied by acidity analyses was performed and described by Armbruster (2000). Moreover, high resolution continuous flow analyses (CFA) was performed for NH4, Ca, liquid conductivity and particle content. Visual stratigraphy by means of digital image processing was done and described by Scholze (1998). Additional microparticle analyses were performed by Saey (1998). 27
Table 3.1: Main data sets for the deep cores KCH and KCS.
Core KCH Core KCS
Analyses Resolution Depth range Resolution Depth range
Stable Isotopes 3.8 – 14 cm 0 – 48.5 m 5.6 – 20 cm 0 – 87.01 m
~ 2 cm 48.5 m – 60.35 m ~ 2 cm 87.01 m – 99.93 m
Major Ions 5 - 10 cm 5 – 10 cm 0 m – 42.39 m 0 m – 28.75 m (Anions and Cations) (Anions and Cations) 0 m – 60.35 m 0 m – 99.93 m (only Anions) (only Anions)
CFA (NH4, Ca, ~ 5 mm 28.75 m - 60.35 m ~ 5 mm 42.39 m - 99.93 m liquid conductivity and particles)
ECM and density 2 mm Complete 2 mm Complete
Visual stratigraphy Complete Complete
3.2 Hydrogen Isotopes
To measure the deuterium content of ice, the melted sample has to be reduced to hydrogen gas. The D/H ratio of hydrogen is then measured with mass spectrometer (MS). Reduction of water to hydrogen at the IUP was done before with hot zinc (Ehhalt 1963). For this work a new sample preparation method using hot chromium instead of zinc (Finnigan “H/Device“) was established. The new method has the following promising advantages (Gehre et al. 1996, Brand et al. 1996): • Good accuracy and precission. • Small memory effect • Due to the short reaction time it can be used as an online method, i.e. the hydrogen gas flows from the H/Device directly into the mass spectrometer. Thus there is no need of storing the hydrogen gas and an autosampler can be used. • Chromium has the potential to form hydrogen gas not only from water but also from other hydrogen-bearing compounds, particularly methane. • A very low amount of sample of less than 1 µl water (or the hydrogen equivalent of other substances) is sufficient for an analyses. This along with the previous item is of interest for the investigation of the carbon cycle at the IUP. 28 3. Methods
3.2.1 Principle of the H/Device The essential components of the sample preparation are shown in Figure 3.2. 1-2 µl of water are injected with a gas-tight microsyringe through a septum into a reaction furnace filled with hot chromium powder. The water vapour formed by quantitative flash evaporation is then reduced (Gehre et al. 1996):
2 Cr + 3 H2O → Cr2O3 + 3 H2 during a certain “reaction time”. The reaction furnace is connected to the sample side of the dual inlet system of the mass spectrometer. When the reaction time is completed, valve 1 at the reaction furnace is opened. Then, during the “equilibration time“, the hydrogen gas is expanded into an “intermediate volume” mainly formed by a stainless steel hose which connects H/Device and MS. The furnace is closed (valve 1), valves 2, 3 and 4 are opened and the gas flows from the intermediate volume into the variable volume of the MS due to the pressure gradient. After a “transfer time“ the variable volume is closed (valve 4), compressed and the measurement by MS can be started. The furnace is evacuated until the next sample is injected during the “evacuation time”.
3.2.2 Adaptation of the method to the MAT 230C mass spectrometer The H/Device is designed to be connected to recent Finnegan mass spectrometer types (MAT 252, DELTA S or DELTA PLUS). Since an older version (MAT 230C) was used for hydrogen analysis, two minor modifications of the H/Device were necessary: 1. The valves which were intended to be computer controlled had to be replaced by manually controlled ones. 2. A first test with the H/Device connected to the MAT 230C showed that the hydrogen pressure in the variable volume was far too low for an accurate analysis of the D/H ratio. The pressure in the variable volume was increased to the required level by reducing the intermediate volume. For that the original steel hose (volume ≈ 92 cm3) was replaced with a much smaller one (volume ≈ 4 cm3). 29
syringe
H/Device
reaction furnace
1 steel hose (intermediate volume) 2
3 mass spectrometer
4 5 sample standard
ion source
Figure 3.2: Sketch of the H/Device for online hydrogen isotope measurements
3.2.3 Effects of apparative parameters The zinc method reduces water samples almost quantitatively. However, δD values of samples prepared with the chromium method are several permille lower than δD values of the same samples reduced using the zinc method. This indicates that the reduction by chromium fractionates the hydrogen isotopes with the deuterium preferably remaining in the reaction furnace. It was investigated in which way this fractionation is affected by the different parameters of the analysis. The results affect the choice of the parameters for the routine analysis in view of a good precision of the method, since variations in these parameters are to some extend inevitable during manual operation. The effect of the various parameters on the measured δD value is shown in Figure 3.3. The not-varied parameters of the respective experiment are listed in Table 3.2. 30 3. Methods
-116.5 a) -116 b)
-117.0 -120 ] ‰ [
-117.5
D -124 δ -118.0 -128
-118.5 1.0 1.5 2.0 0 30 60 90 120 sample amount [µl] reaction time [s]
c) -36 d) -58 -40 ]
‰ -60 [ -44 D δ
-62 -48
700 800 900 0 1 2 3 4 5 temperature [°C] volume ratio
-112 e) -114 f) -114 -116 ] ‰
[ -116
D -118 δ
-118 -120 -120 0 2 4 6 8 10 0 50 100 150 200 equilibration time [min] evacuation time [s]
Figure 3.3: Effects of apparative parameters on the measured δD-value. The absolute values of δD for temperature and volume ratio are different to the others because other samples were used. Volume ratio means volume of reaction furnace over size of intermediate volume. To investigate the variation of temperature the sample was measured repeatedly. Otherwise the results would be biased by the effects of the evacuation time since changing the reaction time takes time. The transfer time had no effect on the measured values. 31
a) Sample amount Sample amount does not critically affect the results. The lower limit is determined by the minimum pressure in the variable volume of the mass spectrometer which allows a measurement. This minimum pressure depends also on the state of the mass spectrometer. b) Reaction time in the furnace Trying reaction times as short as possible and observing the ion currents, practically performed by having valves 1, 2, 3 and 4 open at the injection (see Figure 3.3), indicated that the major part of the water is almost instantly reduced to hydrogen. However, a longer reaction time has the advantage that the measured δ-value is less affected by its variations. c) Reaction temperature The manufacture recommends an optimum reaction temperature of 850°C – 900°C for most compounds. However, for pure water the reduction basically works at temperatures almost down to 500°C. In the standard temperature range the measured δ-value increases only slightly with increasing temperature. d) Size of the intermediate volume The intermediate volume could be varied by expanding to valves 1, 2, 3 or 4. In Figure 3.3 the ratio of volume of reaction furnace over volume of intermediate volume is shown in order to suggest that the volume of different reaction furnaces should also affect the results. e) Equilibration time and f) Evacuation time of the furnace since the last sample Varying equilibrium and evacuation time shows that the H- isotopes escape the furnace faster and that they are also preferably removed during the evacuation step. The values finally chosen for the routine analyses as shown in Table 3.2 are a compromise between time efficiency and the variations during manual operation not being critical for the measured values.
Table 3.2: H/Device parameters for the routine analyses. (1): according to present sensitivity of the spectrometer, of course to be kept constant within one run. (2): Minimum determined by the duration of the mass spectroscopic measurement. (3): proposed by manufacturer
Parameter value for routine analyses
sample amount 1.2 – 1.72 µl (1) temperature of the furnace 850 °C (3) reaction time in the furnace 60 s intermediate volume ~ 6 cm3 equilibration time 30 s transfer time 30 s evacuation time 6.5 min (2) 32 3. Methods
3.2.4 Precision Precision was determined by successive injections of the same sample. The mean standard deviation from five series of each four subsequent injections was 0.26 ‰. This excellent value agrees very well with the value of 0.24 ‰ determined by the manufacturer.
3.2.5 Memory effect Having no memory effect was one of the promised advantages of the chromium method (Brand et al. 1996, Össelmann pers. com). Contrary to that a memory effect of 0.35 ± 0.09 % during 4 tests in February 1998 and 0.94 ± 0.57 % during 5 tests in January 2000 was found (% means permille difference between first and second measurement of the sample per 100 permille difference to the sample before). During all these tests the needle of the syringe was purged several times prior to the injection to remove the water which was potentially remaining in the syringe. The two results are significantly different at a t-test level of 97%. A potential reason for the inconstant memory could be an increasing amount of water remaining in the syringe as the syringe gets worn off. A memory effect probably mainly caused by the syringe is also reported by several other laboratories (Stievenard, pers. com. 1999, Steig pers. com. 1999, Nelson pers. com. 1999). Also different qualities of the chromium powder and the formation of a surface coating from chromium vapour in the intermediate volume may contribute to the memory effect.
3.2.6 Calibration Isotope ratios are measured with the mass spectrometer versus a hydrogen gas working standard (“δprst“) using the double inlet system. These δprst results have to be calibrated with respect to the δ-scale versus V-SMOW (“δSMOW “). As mentioned, the chromium reduction fractionates the hydrogen isotopes and the fractionation depends on the volume of reaction furnace and potentially state of the chromium filling. Since the specifications of furnaces and chromium are not generally constant, a day to day calibration by liquid standards is necessary. For the calibration of the day to day laboratory liquid standards they were measured repeatedly versus the IAEA standards V-SMOW, GISP and SLAP within one run. The δSMOW values of the laboratory standards were calculated from a linear fit through V-SMOW and
SLAP (which define the δSMOW –scale). The accuracy of the fit can be checked by calculating
δSMOW (GISP) from it. It yielded δSMOW (GISP) = 189.5 ± 0.1 ‰ which does excellently agree with the accepted value of –189.73 ‰ (IAEA 1995). The day to day calibration was then performed by measuring 2-3 laboratory liquid standards repeatedly. The δSMOW values of the samples are calculated from a linear fit through the laboratory standards. As samples and reference material are treated identically, this procedure is in full agreement with the “IT- principle”. The importance of this principle was outlined by (Werner and Brand 2001). 33
3.2.7 Standard drift Since the day to day calibration relies on liquid standards, a potential standard drift due to evaporation or exchange with ambient water vapour deserves special attention. The drift rate can be estimated to be about 1 ‰ per 100 measurements from comparing δD of three intentionally frequently opened and measured samples with δD of a backup. Thus for a envisaged maximum drift of 0.3 ‰ after about 30 injections a new aliquot of the standard should be taken from the storage bottle. However, this value refers to 20 ml flasks with wide neck and no care being given to avoid partly evaporated drops getting back in the flasks by the needle of the syringe or the top. In the meanwhile more suitable flasks with narrow neck are used for the working standards and after the respective test the number of injections made by one aliquot of standard may be increased. Alternatively, flasks closed by a septum for the standards could be used.
3.2.8 Overall Reproducibility A standard which is not used for the day to day calibration (“Alpen 2“) was treated and measured as an ordinary sample for 18 months since September 1999. Figure 3.5 shows the respective performance chart.
-88
-90
-92 ] ‰ [
D δ -94
-96
1 Sep 1 Jan 1 May 1 Sep 1 Jan
Figure 3.4: Performance chart for δD values of the “Alpen 2” standard since September 1999. The horizontal line and the grey area indicate average and standard deviation. 34 3. Methods
The average of –93.04 ± 0.7 ‰ fits well to the value determined by direct calibration versus VSMOW and SLAP of –93.2 ‰ . As expected, due to the error in day to day calibration and the memory effect (see section 3.2.6), the standard deviation is larger than for the repeated measurement of one sample (0.26 ‰). The memory effect also contributes to the systematically higher values in June and July 2000 since these results were partly measured in runs of isotopically heavy samples. This evaluation would not show a potential common drift of calibration standards and the Alpen 2 standard. However, since both kinds of aliquots of were renewed frequently enough (see 3.3.7) and independently this can be ruled out.
3.2.9 Summary Measuring δD with the H/Device connected to a MAT 230C MS results in a overall accuracy of 0.7 ‰. This value equals the accuaracy found by other labs (Werner and Brand 2001). It is comparable to the accuracy obtained with the formerly method which uses hot zinc to reduce water. However, the advantage of the new method is that, as it is a online method, a large amount of samples can be measured more efficiently.
3.3 Oxygen isotopes
Oxygen isotopes (δ18O) were measured using the well established method of equilibration with CO2 connected to a MAT 252 MS. the overall accuracy of the method is 0.05 ‰ (Neubert 1998). Thus the accuracy of D-excess values is around 0.8 ‰.
3.4 Stable Isotope intercomparisions
Within ALPCLIM the stable isotope content of cores from the different sites was measured by the laboratories listed in Table 3.3:
Table 3.3: Laboratories involved in ALPCLIM stable water isotope measurements
Laboratory Responsible for site
DISGAM3 Trieste CDL IUP Heidelberg CG LMCE4 Saclay CDL
3 Dipartimento di Scienze Geologiche, Ambientali e Marine
4 Laboratoire de Modélisation du Climat et de l'Environnement 35
Quantifying differences in D-excess requires high accuracy and precision of the stable isotope analyses. In order to be sure that inter-site comparisons are not biased by systematic differences, 13 samples roughly covering the expected range of isotope values were distributed among the labs. Figure... shows the difference of the values from the other labs to the measurements in Heidelberg. The mean δ18O value from DISGAM is about 0.06 ‰ lower than the one from IUP. Neglecting one outlier the average LMCE value is virtually identical to the IUP value. For δD the mean value from DISGAM is about 0.6 ‰ below the again almost equal average values of IUP and LMCE.
DISGAM - IUP 0.2 LMCE - IUP
] 0.1 ‰ [
8
1 0.0 O δ
∆ -0.1
-0.2
-22 -20 -18 -16 -14 -12 -10 δO18 [‰]
3
2
] 1 ‰ [ 0 D δ
∆ -1
-2
-3 -180 -160 -140 -120 -100 -80 -60 δD [‰]
Figure 3.5: Deviations of DISGAM and LMCE stable isotope measurements from IUP values. Since the data from LMCE are not complete it was not appropriate to consider the deviations from the mean. The Figure should not suggest that the IUP values are more reliable then the others.
These differences are among the smallest found in IAEA interlaboratory comparisons. Since the deviations are negligible compared to the variations in isotope signals and since the slight shift of DISGAM values in δD and δ18O is almost compensated in their effect on the D- excess, no correction is needed for any stable isotope inter-site comparisons. 36 4. Glaciological features
4. Glaciological features
4.1 Steady ice flow
4.1.1 Ice deformation
The velocity field of glacier ice (vx, vy, vz) can be inferred from conservation of mass, momentum and internal energy. The respective three equations are intimately coupled through density and temperature. Moreover, a constitutive equation is needed relating applied stresses and ice deformation. This generalised non-linear flow relation for polycristalline ice may be written as (Paterson 1994):
¨ n ε = A⋅τ (4-1) with: A: Flow parameter. It is affected by temperature, density, impurities and ice structure. ⋅ ⋅ ε : Effective strain rate, derived the second invariant of the strain-rate tensor ε ij: ∂v ∂v ε = 1 i + j © ( ) ij ∂ ∂ 2 v j vi τ : Effective deviator stress, derived from the second invariant of the stress deviator tensor. The stress deviator tensor is the stress tensor with the hydrostatic pressure subtracted from the diagonal elements. n: A constant. A value of 3 is applicable for the major part of the glacier volume (Nye 1960). The basic features of the expected velocity field in a glacier as a consequence of the ice flow law are illustrated with a parallel-sided ice slab on inclined bedrock.
x x b z z
α H α H
a) b)
Figure 4.1: Co-ordinate system for the parallel ice slab. a) laminar flow. b) accumulation on the surface First, as shown in Figure 4.1 a), it is assumed that the flowlines are parallel to the surface (“laminar flow”) and that the slab is of infinite extension. The conservation of momentum yield the only non-zero stress component τxz. Then eq. (1) can be analytically integrated with the boundary condition vx(z=H) = 0. This boundary condition reflects the assumption that a cold glacier is frozen to bedrock and thus the ice velocities are zero at the ice-rock transition. 37
The resulting horizontal velocity vx is displayed in Figure 4.2 a). vx decreases rapidly with depth near bedrock, while it is almost constant in the upper part of the glacier.
0 horizontal vertical velocity velocity ) H / z (
h t p e d
e pure shear v i t a l e r
simple shear 1 0 1 0 1 v (z) / v (0) v (z) / v (0) x x z z
Figure 4.2: a) Horizontal velocity (vx) for a 2D infinite parallel slab of ice. b) The respective vertical velocity (vy) for an ice slab with constant accumulation.
The assumption of laminar flow is now abandoned. Instead a constant accumulation rate b on the ice slab is assumed and that the geometry of the slab is stable (Figure 4.1 b)). The horizontal velocities are assumed to be zero at a origin of the co-ordinate system. Additionally a constant shape of the horizontal velocity profile as derived for laminar flow is presupposed.
The magnitude of the vx can be calculated from the conservation of mass: H ⋅ = ⋅ b x ∫ vx (x, z) dz 0
The vertical velocity vz is then determined by conservation of mass: ∂v ∂v x + z = 0 ∂x ∂z and the boundary condition (vz(z=H) = 0). As shown in Figure 4.2 b) the resultant vz decreases approximately linearly in the upper part of the glacier. It approaches zero asymtotically at the base of the glacier: ∂v z = 0 (4-2) ∂ z z=H The latter feature is a consequence only of mass conservation and of the assumption that the glacier is frozen to bedrock. 38 4. Glaciological features
The age at depth z of the glacier is given by: z 1 t(z) = ∫ dz′ (4-3) 0 ′ v z (z ) Thus the vertical velocity profile determines the age-depth relation for ice cores. The typical shape of the vertical velocity profile displayed in Figure 4.2 b) implies that the near bedrock layers contain virtually an infinite time period.
The deformation of ice may be subdivided into the two idealised regimes of simple shear and pure shear as (Figure 4.3). In the upper part of the glacier, where the vx is nearly constant with depth but vz decreases, the ice is subject mainly to pure shear. Contrary to that, simple shear is predominant near the glacier bed, where the change of vz with depth is small but the respective change in vx is at its maximum. A potential indirect relation to stable isotope profiles is outlined in chapter 8.4.
Pure shear Simple shear
Figure 4.3: Deformation of a cubic ice volume under pure shear and simple shear.
4.1.2 Flow models for CG Different approaches were made to develop realistic steady state flow models for Colle Gnifetti. Generally a complex ice flow pattern can be expected due to the irregular bedrock and surface geometry. Moreover, a significant fraction of the glacier consists of firn, which is generally softer than ice and exhibits are more complex constitutive relation (Lüthi 2000). A 2D kinematic flow model using measured surface velocities and a prescribed horizontal velocity profile as shown in Figure 4.2 a) was developed by Haeberli et al. (1988). Numerical finite element models were established by Wagner (1996) and (Lüthi 2000). These models evaluate the stress tensor as being determined by the glacier geometry on a mesh. Stain rates and velocity field are calculated from the flow law. Wagner (1996) considers the effects of firn by modifying the flow parameter A according to the density at a given depth. However, contrary to the generally accepted value of 3, n was taken to be 1.5 in order to obtain a good agreement between measured and modelled surface velocity. Lüthi (2000) implemented a flow law for firn in the finite element code and used n = 3 for the flow law of ice. This so far most sophisticated flow model for Colle Gnifetti was used to 39
calculate backward trajectories for the three deep cores. The results will be used to evaluate systematic upstream effects in chapter 7.2. However, the predicted age-depth relations are critically affected by the assumed bubble close off density. Accordingly experimental and modelled age depth relations for the CG ice cores exhibit significant deviations for the pre- industrial core sections (Armbruster 2000).
4.2 Mass balance
4.2.1 Background For the estimation of systematic upstream effects the mean surface composition upstream the drill site is to be assessed. The mean surface composition is directly related to the long term means of annual accumulation (see Figure 1.4). As outlined below, long term means of annual accumulation can be inferred from the surface mass balance. Furthermore the surface mass balance indicates potential temporal changes of the surface topography. As the surface topography affects annual accumulation and thus the mean surface composition, its potential temporal changes would cause non-atmospheric signals in ice core records.
surface
h2(x,t)
H(x,t)
z h1(x) bedrock
x
Figure 4.4: Coordinate system used for the local surface mass balance
With the coordinate system taken as shown in Figure 4.4, the vertically integrated two dimensional equation of local mass balance may be written as:
∂ h2 (x,t) ∂ h2 (x,t) ∫ ρ ⋅ dz = − ∫ ρ ⋅ v dz + ρ ⋅b (4-4) ∂t ∂x x SNOW h1 (x) h1 (x) with: ρ: density of firn or ice
ρSNOW: density of freshly fallen snow b: accumulation rate 40 4. Glaciological features
Assuming the velocities at bedrock being zero and using the conservation of mass: ∂(ρ ⋅ v ) ∂(ρ ⋅ v ) ∂ρ x + z + = 0 ∂x ∂z ∂t this results in the surface mass balance: ∂H ρ = b ⋅ SNOW − v (4-5) ∂ ρ sub t (h2 ) with the submergence velocity:
∂h v = v (h ) − v (h ) ⋅ (4-6) sub z 2 x 2 ∂x
The submergence velocity measures the downward flow of ice relative to the surface slope. Its geometrical meaning is illustrated in Figure 4.6. The vertical velocity at the ice surface vz(h2) contains downward movement due both, ice flow and potentially seasonally variable densification of firn. The annual accumulation may exhibit a large variability in short term means. However, the submergence velocity reflects the response of glacier flow on the long term mean of annual accumulation. According to eq. (4-5), a discrepancy between accumulation rate and submergence velocity results in a change of surface topography on the respective time scale. Steady state provided, the submergence velocity equals the long term mean of annual accumulation. Then the mean surface composition can be estimated from the submergence velocity at a fixed location. However, also the submergence velocity may exhibit slow temporal changes according to the response time of the glacier.
4.2.2 Mass balance measurement To evaluate the mass balance, surface velocities, surface slopes and accumulation rates have to be measured. For that purpose an array of 16 stakes was installed in September 1998. As shown in Figure 4.5 the arrangement is focused on the upper part of the flowline upstream the drill sites KCS, CC and KCH. In this zone the spatial and temporal variations in surface velocities and accumulation rates are expected to be strong since the surface is strongly curved and wind exposed. The stakes have also served as well known reference points for shallow core sampling and for radio echo soundings. Figure 4.6 illustrates the performance of surface mass balance measurements. During the measuring period (typically one year) the stakes are getting inclined to flow direction mainly in the steeper zones of the surface (Besides, this indicates – contrary to the horizontal velocity for the simple ice slab as shown in Figure 4.2 - a decrease of the horizontal velocity with depth in the uppermost soft firn layers. This feature is contained in the 3D model used for backward trajectories). Since the surface exhibits small scale undulations, the surface slope in 41
ice flow direction was not measured in situ. Instead a ~ 50 m average slope was read from the Swiss digital elevation model for Colle Gnifetti This model is based on the geodetic survey in 1989 at a grid resolution of 20 m (Wagner 1994).
86600 4450 4450 survey stake deep core
KCS
profile 3
CC 86400 KCH profile 2
profile 1
Bergschrund Signalkuppe
86200 633800 634000
Figure 4.5: Stake array to measure surface mass balance. Triangles: stake positions. For profiles 1 and 2: x = 0 m at the intercept with profile 3, x increases towards south-west. For profile 3 x is the curved distance to Signalkuppe along the surface flow line.
Stake positions were measured in September 1998 and September 1999 by geodetic survey using a Leica Wild T 1600 total station with an infra-red distance meter. Following a procedure established by Lüthi (2000), the survey equipment was placed on the glacier surface and its position was determined by using three reflectors which are fixed to surrounding rocks walls. From this defined position a reflector (put on the upper end of each stake) could be surveyed. Stake inclination was measured using a spirit level with a built-in inclinometer and stake declination with a compass. The ice velocity was calculated from the movement of point P (Figure 4.6), where the stake had initially pierced the surface. The overall accuracy of the submergence velocities is around 40 mm/yr. Accumulation at each stake was read in September 1998, September 1999, Mai and August 2000 at an accuracy of ± 5 mm. Due to the sastrugi having a typical height of about 10 cm in the wind exposed low accumulation area, the spatial representativity of the accumulation at 42 4. Glaciological features
single points is lower there. In the area of relatively high accumulation several stakes were about to submerge in September 1999. At these positions new stakes were planted.
Position (t)
Position (t+∆t) Mean surface slope
P
z Surface (t) V V sub x Surface (t+∆t)
Figure 4.6: Measurement of submerging velocities vsub. All velocity data are corrected for stake inclination and refer to the movement of the point P (where the stake had pierced the surface in September 1998).
The measured accumulation rates and submergence velocities are biased by the following effects: 1. Densification and deformation of firn are more rapid at high temperatures. Therefore the vertical velocities and also the horizontal velocities in the steep slopes may be affected from the seasonal variations of firn temperature in the uppermost ~ 10 m (Suter 2001). However, since the velocities shown below are based on two surveys almost exactly one year apart this should not affect the results. 2. Due to the firn compaction the thickness of a firn layer between P and the lower end of the stake in Figure 4.6 is becoming smaller in the course of time. This causes an emerging component for the stake. Thus submergence velocities and accumulation rates are equally underestimated. This effect is expected to be stronger for a stake which is about to submerge (length of a stake is 3 m) than for a newly planted one (planting depth was ~ 1.2 m, which safely protects the stake from getting lost during heavy storms). In fact the stakes planted in 1999 beside the ones planted in 1998 display a systematically higher accumulation rate by about 10 %. To minimise the effect of this process on the results below, always the accumulation rate from the new stake was taken.
3. According to eq. (2), the annual accumulation is to be weighted by ρSNOW /ρ(h2) for the comparison to accumulation rates. However, at all accumulation data refer to strongly wind pressed snow surface. Therefore the increase in density within the respective depth range is small (only around 10% as indicated by the previous item). A 10% correction of annual accumulation will qualitatively not change the findings of the comparison below. 43
4.2.3 Results and discussion The submergence velocities along the profiles 1 – 3 are shown in Figure 4.7. Along the surface flow line (profile 3), submergence velocities are approximately a factor 2 lower in the upper part of the flowline with a discernible minimum around the position of KCH and CC. The uppermost stake in the steep slope above the bergschrund showed a puzzling behaviour: There was zero movement between September 1998 and September 1999. It was submerged in Mai 2000 but it showed up again in August 2000. As it was found tilted, it must have moved. This directly indicates non steady state movement of the steep slope above the bergschrund. Along the two lines perpendicular to the surface flow line (profile 1 and 2) submergence velocities appear to be almost constant toward north-east but increasing towards south-west. This results will be used for the estimation of upstream effects. In order to cover the longest possible period for the assessment of a potential change in surface topography, the submergence velocities are compared to the mean accumulation rate from September 1998 to August 2000. The mean accumulation rates resemble the submergence velocities in the major part of the surface flow line. This indicates that this parts of the glacier was close to the steady state during the two years of observation. Therefore the effects of changing surface topography on the ice core records are expected to be rather small. The main deviations between accumulation rate and submergence velocities occur at the steep slope, at the lowermost point of the surface flowline and south-west end of profile 1. As shown in Figure 1.3, these points are the most proximate to crevasses, suggesting that non steady state conditions at such locations are common. Non-atmospheric signals in ice cores are to be expected particularly at these locations. The variability of the accumulation rates increases in the low accumulation areas and specifically in the steep slope. This is expected to increase the noise in ice core records originating from the respective areas. The negative values of the accumulation rate originate from the period of September 1999 to Mai 2000, when probably the hurricane “Lothar” at christmas 1999 eroded up to 35 cm compact firn in the most wind exposed areas. Such events limit the accuracy of annual layer counting for dating. 44 4. Glaciological features
profile 1 profile 2
3 submergence velocity 3 range of accumulation rate mean accumulation rate KCH 2 2 ] a / m [
b
u
s 1 1 v
, b
0 0
-1 -1 0 100 0 100 position [m] position [m]
profile 3
3 KCS Bergschrund CC
2 KCH ] a / m [
b
u 1 s v
, b
0
-1 0 100 200 300 400 100
10 b / b
∆ 1
0.1 0 100 200 300 400 position [m]
Figure 4.7: Surface mass balance for the three profiles as shown in Figure 4.5. Submergence velocity (dots), mean accumulation rate from September 1998 to August 2000 (squares) and range (minimum and maximum) of accumulation rates (dashed lines,based on the observations from September 1998, September 1999, Mai 2000 and September 2000). The lowermost graph shows the relative variation of the accumulation rate, i.e range(b)/mean(b). 45
5. Dating
5.1 State of the dating
High elevation alpine glaciers are characterised by small scale geometry and strong spatial variations in annual accumulation. Therefore down to bedrock cores from such drill sites exhibit a strongly non-linear age-depth relation. This requires the combination of different dating methods: • Annual layer counting • Identification of absolute time markers. The horizons of fallout from nuclear weapon tests are unambiguous time markers only back until 1954. A much larger time range is covered by fallout horizons from volcano eruptions and Saharan dust events. 18 • Records of long lived atmospheric trace gas concentrations (mainly δ O in O2 and methane) can be established from air bubbles in glacier ice. The records from Alpine ice cores can be matched to well dated records from polar ice cores. • Radioactive decay methods Age-depth relations based on annual layer counting and absolute time markers were established for KCH, KCS (Armbruster 2000) and CC (Schäfer 1995). However, the proposed age-depth relations were based on different data sets and are thus inhomogeneous concerning both time range and accuracy. Age constraints for the basal layer sections are separately discussed in chapter 8. For dated sequences the following starting position was encountered:
KCS Volcanic horizons, characterised by sulfate concentration being 2⋅σ above mean level accompanied by an enhanced acidity, were identified almost throughout the complete pre- industrial core section. Continuous annual layer counting using the NH4 high resolution CFA profile confirmed by unambiguous time markers was possible down to a depth of zo = 45.1 m
(Laki at time t0 = 1783 A.D.). Below this depth the error in annual layer counting caused by core sections of poor seasonal signal quality gets too large. Therefore, below this depth the mean annual layer thickness λ(z) (for steady state equalling the vertical velocity vz(z)) was determined for core sections with well visible seasonal variations. An upper limit v+(z), a best guess v0(z) and a lower limit v-(z) for the vertical velocity over depth were established by fitting the data points. This was possible down to a depth of ≈ 62 m w.e. 46 5. Dating
relative depth
0.0 0.5 1.0
20
10
KCH 0 0 10 20 30 40
1809
20 e r o c
] f o m
c [ d
10 n λ e CC 0 0 10 20 30 40 60
40
1784 20 KCS 0 0 10 20 30 40 50 60 70 80 depth [m w.e.]
Figure 5.1: Mean annual layer thickness over depth between dating markers for KCH, CC and KCS.: Dotted for CC: according to the dating hypotheses of Schäfer (1995). Solid: Corrected version (this thesis, see below).
Next a time window [t-(zn) ... t+(zn)] was calculated for the n-th identified volcano horizon at depth zn using = − zn 1 t+ − (z ) t ∫ dz´ (5-1) , n 0 z 0 v+,− (z´)
Additionally a sequence of volcano eruptions potentially recorded at Colle Gnifetti was established from historical and polar ice core records. One of the eruptions in the sequence 47
being within the time window [t-(zn) ... t+(zn)] can then be assigned to the n-th ice core volcano horizon. With this procedure a volcano chronology back to Elgja (A.D. 934) in a depth of 61.41 m w.e. has been proposed with a maximum error of 50 years (Armbruster 2000). However, a closer look at the proposed volcano chronology reveals that the resultant mean annual layer thickness between volcano horizons (zoomed view in Figure 5.2) turns out to exhibit a slight transient increase with depth in some core sections before 1784 A.D. Basically it is indeed possible that the annual layer thickness increases with depth due to upstream effects, temporally variable accumulation rate or the potential occurrence of compressive flow upstream of bedrock boulders. It could, however, also be a hint on a non-perfect volcano chronology.
15
10 ] m
c 1784 [
λ
5
KCS 0 40 50 60 depth [m w.e.]
Figure 5.2: Zoomed view of annual layer thickness over depth for KCS. Dotted: following the dating hypotheses of Armbruster (2000). Solid: Alternative version, see below.
The increasing annual layer thickness in sections could be caused by three problems: • First, for all volcano horizons the time window was calculated from the same starting
depth z0. This causes large time windows in the lower core sections as the error in vz(z)
between z0 and z is accumulated. If then a certain volcano horizon is assigned to an eruption at the younger limit of its time window but the next to an event at its older limit an irregular annual layer thickness results. • Second, the sequence of historical volcano eruptions with the potential for volcanic fallout at Colle Gnifetti was not complete both with respect to the most proximate sources (Italian volcanoes, Canary Islands and Azores) as well as the eruptions recorded in polar ice cores. 48 5. Dating
The record of the proximate volcanoes will probably never be complete due to the ceasing historical records several centuries ago. • Third, it was considered as a must to assign an eruption to each volcano horizon, which was generally not possible for polar ice cores (Bigler et al. 2001).
KCH The dating procedure was supposed to be analogous to KCS. Due to the lower accumulation rate, continuous annual layer counting based on the NH4 profile could be performed only down to the Saharian dust event in 1901 at a depth of 20.1 m w.e. Annual layer counting in special sections was possible down to a depth of ≈ 40 m w.e. However, due to uncompleted major ion and acidity profiles, so far no volcano chronology was set up for KCH. Therefore just a preliminary age-depth relation tλ(z) solely based on fits of vz(z) was established back to ≈ 1300 A.D. with an error of ≈100 years.
CC Contrary to the two cores drilled in 1995, no high resolution CFA profiles were performed at the CC core. Instead discrete samples were measured by ion chromatography in 2 cm resolution. Therefore the identification of seasonal variations based on the NH4 record was possible only back to the volcano horizon of Tambora 1815 at a depth of 27 m w.e. Below this depth an age-depth relation based on the three most prominent volcano horizons and the pre-dating by a 3-D flow model by Wagner (1994) back to 1631 A.D. at a depth of 38.8 m w.e. was proposed by Schäfer (1995). However as illustrated in Figure 5.2, the mean annual layer thickness between dated horizons is far from decreasing monotonously with depth below the 1809 horizon. This is clear evidence that the lower part of the CC core is much older than stated by existing age-depth relation. Besides, it was noted that for fitting continuous age-depth relations a major visible dust layer was not accurately assigned to 1936 or 1937 and that the unambiguous tritium maximum of 1963 had not been considered at all. An additional problem is the identification of volcanic layers. Several of the volcanic horizons for KCS, identified by an enhanced sulfate concentration accompanied by high acidity, exhibit a virtually zero ECM signal. This is contrary to all expectations (Hammer 1980). As a high ECM signal was a major criterion for the selection of volcano horizons in CC, also the existing set of volcano layers had to be re-evaluated.
5.2 Progress during this work
From the situation as described above, the following steps were initially intended: • Refine the volcano chronology for KCS such that an increase of λ(z) with depth z does not occur. 49
• Establish an age-depth relation for KCH before 1900 by identifying volcano horizons and assigning them to historical eruptions with the constraints from λ(z). • Re-evaluate the CC volcano chronology completely • Identify other common horizons, like Saharan dust layers in order to establish well synchronised time series.
KCS The set of volcano eruptions since 876 A.D. with potential impact on Colle Gnifetti was slightly modified. As shown in Annex B the volcanoes are subdivided in three categories: Bi-hemispheric events (B-event). Such events are recorded both in Greenland and Antarctic ice cores. It may be assumed that without exception all Bi-hemispheric events during the 19th century (unknown 1809, Tambora 1815, Coseguina 1836 and Krakatao 1883) were detected in the already well dated sections of the Colle Gnifetti cores. Thus it is reasonable to assume that also the Bi-hemispheric events before caused most probably a detectable layer of volcanic fallout at Colle Gnifetti. However, these layers will be harder to detect with increasing depth due to annual layer thinning. Northern hemispheric events (N-event), recorded only in Greenland ice cores. The well dated sections suggest that these events have a much lower probability to show up in the CG cores. Local eruptions (L-event) of volcanoes in Italy, Canary Islands and Azores (Simkin and Siebert 1994). Only moderate and large events characterised by VEI5 ≥ 3 are listed in Annex B and taken into consideration as potential time markers. It is hard to tell whether also the numerous minor events (VEI ≤ 2) may have caused a signal at CG. The events several centuries ago are probably all documented. The events were successively assigned to the identified volcano horizons as follows: The time window for the (n+1)-th volcano horizon was calculated (not from the 1st as done by Armbruster) but from the n-th horizon, which had been already assigned to a historical eruption, as starting point:
= − zn+1 1 t+ − (z + ) t ∫ dz´ (5-2) , n 1 n z λ n +,− (z´)
The range λ+ ... λ- according to Armbruster (2000) is too large and was replaced with a more conservative estimation.
5 VEI: Volcanic explosivity index. It characterises the strength of an eruption on a scale of 1-8. 50 5. Dating
Table 5.1: Refined volcano chronology for KCS.
depth time window year events within time window [m w.e.]
45.1 Starting point 1784 Laki 1784 45.52 1769 – 1770 1770 L: Vesuv 1770-1779, L: Vulcano 1771 46.69 1728 – 1731 1731 N: Orafajokull, Iceland 1728, N: Lanzarote 1731, L: Vulcano 1727, 1731 47.37 1701 – 1708 1707 L: Vesuv 1706-1707 48.65 1646 – 1664 1660 N: Long Island 1660, L: Vesuv 1654-1680: L: Vulcano 1651, 50.17 1583 – 1613 1600 B: Huaynaputina 1600, N: Billy Mitchel 1588, N: Kelut 1589, B?:Ruiz, Columbia 1595, N: Momotombo 1605, L: Etna 1595 (?) 51.57 1530 – 1549 1550 L: Vulcano 1550, L: Ätna 1536 53.82 1407-1462 1460 B: Kuwae 1460,N: unknown 1403, N: unknown 1441, L: Azores 1444(?), L: Vulcano 1444 54.97 1390 - 1409 1396 N: unknown 1403, L: Tenerife1396, L: Azores 1400 L: Etna 1408 55.12 1381-1402 1389 L: Etna 1381(?), N: Hekla 1389, L: Tenerife 1396 57.08 1249-1312 1260 B: El Chicon 1260, B?: unknown 1285 L: Ischia 1302 61.41 896- 1038 943 N: 895,901.903,916,932,937,939,940,943,1027
If an Bi-hemispheric event was found within the (n+1)-th time window it was assigned to this horizon only if this caused no pronounced increase in v(z) with depth. If no Bi-hemispheric event fulfilled this conditions, the time window for the (n+2)-th, (n+3)-th... volcano horizon was calculated until a Bi-hemispheric event was found. Thus the most prominent Bi- hemispheric events (1600 A.D., 1450 A.D. and 1260 A.D.) were unambiguously assigned to KCS volcanic layers (Table 5.1). The “skipped” volcano horizons were assigned to the more numerous Northern hemispheric and Local events with the constraint of a decreasing annual layer thickness. However, the result turns out to be a minor modification of Armbruster (2000).
KCH A part of IC samples from KCH could not be measured right after they were filled in vials. Therefore the vials were sealed in PE bags and stored in the cold room at –20 °C for several months. Unfortunately this affected the acidity of the samples, as they either showed a 51
generally increased acidity level or the acidity analysis could not be evaluated (Ruth pers. com). Only two horizons at depths of 35.71 m w.e. and 41.51 m w.e. could be identified as volcanic layers. However, the event at 41.51 m w.e. is below the sections where annual layers can be detected. Moreover, the sulfate concentration of ≈ 1300 ppb is much higher than for all other volcanic layers. This causes some doubt on its volcanic origin. Other potential reasons for this signal are discussed in section 8.4.
Table 5.2. Saharan dust events and volcano layers identified in KCH and time range estimated from annual layer thickness, integrated with the previous event as starting point. 1: Historical date. 2:Age taken from KCS chronology 3: not detected in KCS due to missing data.
depth time window year events within time window [m w.e.]
20.12 starting point 1901 Saharan dust 1901 25.6 1840 – 1867 1863 Saharan dust 1863 (1), (3) 29.9 1796 - 1831 1830 Saharan dust 1830 (1)
34.1 1723 – 1778 1750 Saharan dust ≈ 1750 (2) 35.71 1685 – 1716 1707 Vesuvius 1706-1707 (2)
38.4 1552 – 1634 1627 Saharan dust ≈ 1590 (2)
Saharan dust ≈ 1627 (2)
As no volcano chronology can be established on the base of one horizon, the four visible dust layers before 1900 (shown in Figure 5.3) were used as age markers as follows: With the estimated vertical velocities a time windows according to eq. 5.2. was calculated. Like for
KCS the estimated range v- ... v+ was considerably increased. Otherwise not even the well known 20th century time markers are within the respective time windows. The layers in KCH can then be assigned to Saharan dust or volcano events from the KCS chronology which are within in the time window. Additional constraint was that also for KCH the annual layer thickness decreases monotonously with depth. As a result there is only one possibility to match KCH and KCS horizons, except for the lowermost one. The result is shown in Table 5.2. For the last horizon 1627 A.D. was accepted as otherwise the matching of KCH and CC would result in an increasing annual layer thickness in a section of CC.
CC From the set of volcano layers according to Schäfer (1995) almost 40 % were not taken into consideration as they do not meet the condition of showing a significantly enhanced sulfate 52 5. Dating
level accompanied by high acidity. One new layer was added. The corrected set of layers is listed in Table 5.3:
Table 5.3: Revised volcano chronology for the CC core. 1: Properly assigned by Schäfer (1995). 2: Event assigned by matching with KCH and KCS
depth [m w.e.] event
14.75 Katmai 1912 (1) 19.19 Krakatau 1883 (1) 27.25 Tambora 1815 (1) 31.11 Vesuvius 1707 (2) 37.52 Kuwae 1460 (2) 38.08 Tenerife 1396 or Etna 1381 (2) 38.8 El Chicon 1260 (2)
These volcano layers as well as major Saharan dust events below Tambora 1815 had to be matched to the records of KCH and KCS. Figure 5.3 display an overview of major markers in the lower core sections. Again it was considered as a must that the annual layer thickness decreases monotonously with depth. The matching was done as follows: As shown in Figure 5.3, the distances between corresponding events in KCS and CC in the well dated section above the Saharian dust event 1830 closely correspond. Above a depth of ≈ 30 m (1890) this correspondence of the layers in KCH and CC is confirmed by the isochronous internal layers detected in a Radar profile along the flowline which was performed in October 2000 by Uwe Nixdorf, AWI (Figure 5.4). 53
0.2 7 0 2 1 3 6 8 1
1
~ t
t 0.1 s s u u d d
0.0 particles: no data KCH 0 0.2 20 30 40 ]
. 1 e 0 n . 0 o w 8
z i ~ r
m 0.1 t [ o
s h s
u r o d e n y a a
l 0.0 c
l t CC o s v
u 0 d 0.2 20 30 40 1
0.1
particles: particles: 0.0 no data no data KCS 0 40 50 60 70 80 depth [m w.e.]
Figure 5.3: Visible Saharan dust layers (square with solid line), additional Saharan dust layers identified by major peaks in CFA particle profile for KCS (square with dotted line) and volcano layers (yellow) in the lower core sections. Arrows indicate matched horizons. Green: Lower limit of age constraint from annual layer counting
Assuming that this similarity of distances still holds down to the last dated event in KCH, the CC layers down to the dust layer in a depth of 33.1 m w.e. (1627 A.D.) can be matched to KCH. Assuming that the Bi-hemispheric volcanic events 1260 A.D. and 1460 A.D. were also recorded by CC, the remaining three volcano horizon of CC were matched to the 1460, 1389 and 1260 event in KCS. This is the “youngest” combination that causes no increase of annual layer thickness for CC. However, this is not unambiguous as the two dust layers in CC at 31.9 m w.e. and 43.66 m w.e. do not have a corresponding layer in KCS. Also the matching of the last visible dust layer is just a tentative suggestion. Alternatively one could match the dust layers of KCS and CC. However, then the volcanic layers do not fit. Although the sulfate concentrations of the three lowermost volcanic layers of CC are much lower than for the strange lowermost volcanic layer in KCH (see p. 45), an artifact may not be a priori excluded. Anyway, the resultant annual layer thickness for CC is shown in Figure 5.1. The strong 54 5. Dating
decrease in annual layer thickness below 27 m w.e. is remarkable. This feature is also seen for the well dated KCS core below a depth of 37 m w.e. Even for KCS this feature can only partly be explained by an upstream effect. The annual layer thickness for all three cores seems to be in good agreement with the theoretical expectation of approaching zero asymtotically at bedrock (eq. 4-3).
Figure 5.4. (Nixdorf, pers. com.): Radar profile along the surface flow line (Figure 1.3). Frequency was 250 MHz. Glacier surface through the top of the three cores and bedrock at the bottom are clearly visible. The indicated horizons of 1920 and 1890 were estimated with a high resolution large scale plot.
5.3 Summary
For KCS, a volcano chronology with the constraints of annual layer counting back to 943 A.D. was established. The maximum uncertainty is increasing to around 50 years at the lowermost horizon. For KCH, Saharan dust layers and one volcanic horizon were matched to the KCS chronology back until 1627 within the constraints from annual layer counting. The maximum error is estimated to around 15 years back until 1707 and up to 30 years below. For CC, the horizons above 1627 were matched to KCH and KCS. Provided that the matching is true, the uncertainty in this section equals the uncertainty for KCH. The volcanic layers below were matched to volcanic horizons of KCS with the constraint of a decreasing annual layer thickness with depth. However, this is to be considered just as a best guess. 55
The dating of the CG cores could be improved by the following activities: • With a high resolution sulfate and acidity profile performed by CFA it could be possible to identify more volcano layers. The estimation of annual layer thickness may be improved by a multi-parameter approach. • Specific volcano eruptions could be properly identified by comparing the tephra in CG cores to other sites. • A Radar profile along the surface flowline with a lower frequency, e.g. 100 MHz could be measured. As radar waves of lower frequency can penetrate the glacier ice to a larger depth, however at expense of resolution, isochronous layers in the uncertainly matched lower core sections may be identified. Thus the three records may be synchronised. • By establishing a method for 14C (DOC) an age constraint for the basal layer may be obtained. This work is in progress. 56 6. Sampling trait
6. Sampling trait
An ice core at the upper Grenzgletscher (4200 m asl., 1 km from Colle Gnifetti at a wind protected site) exhibits an annual accumulation of 2.7 m w.e./yr (Eichler et al. 2000). The annual precipitation at Colle Gnifetti should be comparable to this value. However, the annual accumulation in Colle Gnifetti long term records is only 0.24 m w.e/yr – 0.52 m w.e/yr and therefore smaller by a factor 5 - 10. This small finally conserved fraction of annual precipitation is strongly selective with respect to season and weather type. An assessment of the discrimination is mandatory for the evaluation of inherent stable isotope signals.
6.1 Recorded season
6.1.1 Melt events Dry deposited snow gets protected from wind erosion by melt events leading to ice layers, by the formation of ice crusts and by intense sintering of firn. Alean et al. (1983) suggested that melt layers control the annual accumulation at CG. Melt events occur if the snow surface temperature reaches 0°C. Laboratory studies on CG firn samples (Beck 1985) demonstrated that melt events are favoured by: • high diurnal air temperature • intense incoming radiation • low wind speed (if the air temperature is below the snow surface temperature) • high relative humidity The study pointed out that melt events occur even at air temperatures well below 0°C under calm conditions, whereas wind prevents surface melting at positive air temperatures (Beck et al. 1988). However, not only the present values of the relevant parameters, but also their temporal evolution determines whether melt events occurs or not (Beck 1985). Therefore, modelling effort would be necessary to evaluate site specific meteorological time series. Additional problems are the feedback of surface melting on the albedo of snow (Paterson 1994) and the influence of snow impurities on the albedo (Warren 1980). However, the mean seasonal variations of melt events at CG can also be empirically estimated from the observed number of melt events per month at the Seserjoch energy balance station, situated 1 km from CG at 4300 m asl. At this site air temperatures, snow surface temperatures, radiation fluxes and wind speed were measured from May 1999 – April 2000 within ALPCLIM (Suter pers. com.). Table 6.1 displays the number of melt days per month, the monthly means of air temperature Ta and of incoming total radiation R at Seserjoch in Summer 1999. Wind speed was not taken into consideration due to data gaps. Moreover, wind speed is expected to exhibit a less pronounced seasonal cycle compared to temperature and radiation. 57
Table 6.1: Melt days (characterised by a diurnal maximum of snow surface temperature above –1°C), monthly means of air temperature Ta and monthly means of incoming radiation R at Seserjoch (Suter, pers. com.). During all other months until April 2000 no melt events occurred.
2 Month Ta [°C] R [W/m ] Melt days Melt days (observed) (fitted)
May 1999 -8.3 273 5 3.2 June 1999 -7.1 309 4 6.8 July 1999 -4.3 321 22 19.8 August 1999 -4.1 257 18 18.6 September 1999 -5.5 211 5 7.3
A non-linear double parameter fit was established to calculate the number of melt days from
Ta and R. As shown in Table 6.1, the fit function reflects the decreasing number of melt days from July to September where the air temperature is almost constant but the incoming radiation decreases. Besides it yields virtually zero melt days for all other months until April 2000 congruent to the observations.
-5 radiation temperature 300 ] 2
-10 m / W [
200 n o i ] t a C i ° [ d
a T
-15 r
g n 100 i m o c n i
-20
0 1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 6.1: Seasonal variation of air temperature expected at 4500 m asl (Böhm 1999) and seasonal variation of incoming solar radiation at Pian Rosa, Monte Rosa Massiv, ≈ 4000 m asl., 1962-1992 (data source: NASA; http://harp.gsfc.nasa.gov/~imswww/pub/imswelcome/plain.html).
To estimate the mean seasonal variation of melt events at CG, the fit was evaluated with (Figure 6.1):
Ta : expected mean seasonal variation of air temperatures at 4500 m asl. R : the mean seasonal variation of incoming radiation in the Monte Rosa massif (on a flat surface) 58 6. Sampling trait
As shown in Figure 6.1, the temperature maximum is at the end of July, while the maximum of the incoming radiation is in June.
more sunny flat surface 10 more shadily s y a d
5 t l e m
0
1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 6.2: Estimated seasonal variations in number of melt events per month at Colle Gnifetti.
The result for the estimated number of melt days per month at Colle Gnifetti is shown as “flat surface” (dots) in Figure 6.2. However, the incoming total radiation R contains a contribution of direct solar radiation which strongly depends on the position at Colle Gnifetti (Alean et al. 1983). Therefore, as shown in Figure 6.2, for a more sunny location (R increased by 10%) the seasonal maximum in melt days is rather in June and not in July as it is for the flat surface and more shadily (R reduced by 10%) places. The calculated potential direct solar radiation in the Monte Rosa massif exhibit a large range in annual means of 0 – 25 W/m2 (Suter 2000). However, since the value for Seserjoch (22 W/m2) resembles the respective values at the drill sites (19-20 W/m2), the mean seasonal variations in melt days according to Figure 6.2 are expected to be a reasonable approximation.
6.1.2 Wind erosion The effect of wind erosion on the seasonal distribution of the finally conserved precipitation is estimated by a simple statistical model. It assumes:
• The probability of melt events (per day) during the i-th month mi is adapted from the number of melt days as shown in Figure 6.2. • The probability of wind erosion (per day) e denotes the probability that all precipitation events since the last melt event are eroded completely. e may as well exhibit a seasonal cycle but because of the lack of reliable site-specific long term records it was assumed to be a seasonally constant parameter. 59
• The probability of precipitation events (per day) p was assumed to be seasonally constant as well. • A precipitation event is counted as “conserved” if the first sequencing melt event occurs before the first sequencing erosion event. Thus the finally conserved fraction of
precipitation events ci for the i-th month can be evaluated.
1.0 ) i 0 f c ( o
0.01 s n t o
n 0.1 i t e c
v 0.2 a e r
f 0.5
n d o 0.5 i e t v a r t i e p i s c n e o r c p
0.0 1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 6.3: Modelled seasonal variations of the conserved fraction of precipitation events at CG with the probability of wind erosion, e, as parameter.
To calculate the long term means of ci, the occurrence of precipitation-, melt- and erosion events was simulated by random generator over 1000 years. Figure 6.3 shows the resulting modelled conserved fraction of precipitation events for various values of the probability for wind erosion. As expected, the seasonal cycle of ci closely resembles the seasonal variation of melt events for intense wind erosion. As expected, the main increase in the conserved fraction of precipitation events for weaker wind erosion occurs in spring, since the period until the subsequent melt events is shorter.
6.1.3 Seasonal variations in precipitation amount Precipitation amount and its seasonal timing exhibits a strong spatial variability in the Alps. The total annual precipitation strongly increases with altitude. However, the seasonal variation of precipitation amount normalised (such that the sum is one) are not expected to change a lot with altitude. This normalised precipitation amount of the i-th month is denoted as ni. The typical seasonal variation of precipitation amount in the Monte Rosa massif displayed in Figure 6.4 are characterised by a pronounced maximum in spring while mid- summer and mid-winter are relatively dry. 60 6. Sampling trait
150
] 100 m m [
n o i t a t i p i
c 50 e r p
0 1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 6.4 Mean seasonal variation of precipitation amount in the Monte Rosa massif based on the precipitation series from Diga Gabiet 1928-1999 (located at the south slope at 2540 m asl.).
The seasonal distribution of the finally conserved apportion of snow is now assessed from the normalised seasonal variation of precipitation amount ni weighted with the conserved fraction of precipitation events ci . With ai being the apportion of precipitation from the i-th month: c ⋅ n a = i i i ⋅ ∑ci ni year
The apportions ai are shown in Figure 6.5 for various probabilities of erosion. As expected, the seasonal composition of cores from sites with strong wind erosion resembles the seasonal variations of melt events, while for the cores from sites with weak wind erosion it reflects the seasonal variation of precipitation amount. For strong wind erosion, the estimated monthly apportions show hardly any change with the degree of wind erosion. 61
25 0 0.01 0.1
] 20 0.2 (KCS) %
[ 0.5 (CC or KCH)
n o i t r 15 o p p a
l
a 10 n o s a e s 5
0 1 2 3 4 5 6 7 8 9 10 11 12 month
Figure 6.5: Mean seasonal distribution of precipitation amount in ice. Number categories refer to hypothetical probabilities of wind erosion
6.1.4 Application on ice core records The result of Figure 6.5 is now linked to the observed (net) annual accumulation and winter snow content of CG ice cores.
Accumulation rate The annual accumulation b of the core is related to the total annual precipitation N at the sampling site by: = ⋅ ⋅ b N ∑ pi (e) ni year
Relation 6.2 was evaluated with N = 2.7 m w.e./yr (from the core at the upper Grenzgletscher core) with the erosion probability as parameter. Figure 6.6 shows the relation between net annual accumulation and probability of wind erosion. 62 6. Sampling trait
] 3 r y / . e . w
m [
b
n 2 o i t a l u m u c c a
l
a 1 u n n a
t KCS e n CC, KCH 0 0.0 0.1 0.2 0.3 0.4 0.5 erosion probability [1/d]
Figure 6.6: Solid: Relation between erosion probability and net annual accumulation for a total annual accumulation N = 2.7 m w.e./yr. Dotted: for N = 2.2 m w.e./yr and N = 3.2 m w.e./yr.
Figure 6.6 indicates that the annual accumulation of KCS (b = 0.52 m w.e./yr) corresponds to an erosion probability of ≈ 0.2 and the annual accumulation of KCH and CC (b = 0.25 m w.e./yr) to an erosion probability of ≈ 0.45. According to Figure 6.6, these values do not critically depend on the assumed total annual precipitation. With this values for e the seasonal distribution of precipitation of the respective core can finally be inferred from Figure 6.5. According to Figure 6.5 there is hardly any difference in the seasonal distribution of precipitation amount for KCH, CC and KCS.
Winter snow content
From all major ions, NH4 exhibits the largest seasonal variation both, in air and in snow concentrations at high elevation Alpine sites. The seasonal variation of NH4 snow concentrations from a high accumulation core at CDD is displayed in Figure 6.7 (Preunkert et al. 2000). The seasonal variation is characterised by a bisection of the year in a - each almost constant - winter and summer level. This enables the estimation of the winter snow content in
Alpine cores through selecting the samples of NH4 winter levels. Measurements of NH4 air concentrations at CDD confirm that the onset of high NH4 levels is around the beginning of April, while the winter level start at the beginning of October (Preunkert et al. 2001). 63
1000
] 100 b p p [
4 H N
10
1 2 3 4 5 6 7 8 9 10 11 12 13 portion of the year
Figure 6.7: Seasonal variations of NH4 snow concentrations from a high accumulation core at CDD adapted from Preunkert et al. (2000). Portion of the year was inferred from the matching seasonal variations in stable isotopes and instrumental temperatures. Mind the logarithmic scale.
The winter snow content of the core W is related to the erosion probabilities by:
⋅ ∑ci (e) ni W = wint er ⋅ ∑ci (e) ni year
The graph of this relation is shown in Figure 6.8. As shown in Figure 6.8 the relation between winter snow content and erosion probability is critically affected by the assumed beginning of
NH4-summer levels. This is due to the almost simultaneous onset of snow conservation for low accumulation cores (Figure 6.5) and the beginning of NH4 summer levels. However for a winter snow of 10% - 20 %, as found for the low accumulation cores, the erosion probability is above ≈ 0.1. Thus, according to Figure 6.5, the corresponding seasonal distribution of the conserved snow is certainly very similar to the estimations from the net accumulation rates. 64 6. Sampling trait
60
50 ]
% 40 [
w o n s 30 r e t n i w 20
10
0 0.0 0.1 0.2 0.3 0.4 0.5 erosion probability [1/d]
Figure 6.8: Solid: Relation between winter snow content and erosion probability for an onsets of the
NH4 summer at April 1st. Dotted: for March 15st and April 15st.
6.2 Recorded weather types
6.2.1 Precipitation amount Synoptic weather types in the western Alps can be classified according to the scheme of Schüpp (Fliri and Schüpp 1984). This scheme is based on surface pressure and 500 hPa heights within a circle of 444 km diameter covering the western Alps. If the wind field is homogeneous, a day is classified as advective weather type, otherwise as convective. Advective weather types are subdivided according to the overall direction of circulation (N, NW,...), convective types in high (H), flat (F), saddle (X) or low (L) pressure distribution. A detailed description of the scheme is found in Fliri and Schüpp (1984). Figure 6.9 displays the percentage of annual precipitation at different weather types for Monte Rosa region (MR), Mont Blanc (MB) massif and the Bernese Alps (BA). According to the location at the southern slope of the Alps, the percentage of annual precipitation at different weather types in the Monte Rosa region is strongly concentrated on the advection from south- west (SW). This distribution is very variable within the Alps e.g. the main amount of precipitation in the Mont Blanc Region occurs at W, while in the Bernese Alps it occurs at the weather types W, SW and L. According to Auer (1999), the distribution of precipitation at different weather types shows no significant change with altitude in the Monte Rosa massif. 65
N MR 30 MB NW NE BA frequency 20 n o i t a
t 10 i p i c e r 0 W E p
l a u n 10 n a
f o 20 %
SW SE 30
S
Figure 6.9: (Auer pers. com) Percentage of annual precipitation at different weather types in the region of Monte Rosa (MR), Mont Blanc (MB) and the Bernese Alps (BA). Dashed: frequency of the circulation type.
As expected from Figure 6.9, the number of weather types associated with a amount of precipitation above average in is less for the MR and MB region than for the Bernese Alps (Fliri and Schüpp 1984). This could cause an increased sensitivity of precipitation amount with respect to frequency of weather type.
6.2.2 Fresh snow samples Twenty-five fresh snow samples were collected at the summit of Signalkuppe during the summer seasons 1998 – 2000 by the CAI6-staff of the Capanna Regina Margerita. These samples cover the period middle of June to middle of September. Four additional samples were collected in winter 2001 at the Klein Matterhorn (3815 m asl, 11 km west from Colle Gnifetti) by the cable car staff. Weather conditions (storm or fog), amount of precipitation, precipitation type (snow flakes or riming event) and air temperature were recorded at the time of sampling. The synoptic weather type prevailing at the precipitation event according to the classification of Schüpp was kindly made available by Stephan Bader, Meteoschweiz. Only very low correlation of both δ18O and D-excess, to the occurrence of storm, fog, riming, and to the amount of precipitation were observed. A low but positive correlation of δ18O with temperature was observed.
6 Club Alpino Italiano 66 6. Sampling trait
*
20 ]
‰ 15 S [
s SW s
e W c
x NW e - N D 10 L * F * E NE * 5
-30 -20 -10 0 δ18O [‰]
Figure 6.10: Connection of δ18O and D-excess of fresh snow samples and weather type. *: Winter snow from Klein Matterhorn. Weather type data: Bader, Meteoschweiz (pers.com).
Figure 6.10 shows the slight positive correlation of δ18O and D-excess for the summer samples. This corresponds to a slope ∆δD/∆δ18O (obtained by linear regression) of 8.25 ± 0.1. It is a common finding that the slope of ∆δD/∆δ18O in precipitation may be different from 8 in specific seasons (Rozanski et al. 1982). Agemar et al. (2000) found a slope of almost 10 for Heidelberg water vapour in July. Taylor (1968) found the slope in vapour to increase with altitude in water vapour. This was attributed to the adiabatic cooling history. The outlier of high D-excess occurred at the lowest observed temperature of –19°C. In fact, a high D-excess is expected if an air parcel encounters vapour-solid condensation low temperatures. Within most weather types the range of the D-excess more than 6 ‰. This and the average value being well above the low altitude summer lever may indicate that the D-excess depends not only on the source region conditions of the water vapour but may also be affected by the local condensation process. Despite the low number of samples, the δ18O means for the four most frequent weather types in summer as listed in Table 6.2 are significantly different. Advection of air masses from northern directions (NE, N or NW) is accompanied by δ18O being roughly 10 ‰ lower than in the remaining samples, while these weather types do not cause a specifically low temperature. Since these air masses had to cross the main topographic divides of the Alps to reach the sampling site accompanied by intense loss of precipitation at the windward slopes this is probably mainly an amount effect. 67
Table 6.2: Typical δ18O values for different weather types in summer. (1): weighted by approximate precipitation amount
weather type number of events mean δ18O [‰]
F 3 -9.1 ± 1.4
SW 13 -9.5 ± 2.9
W 3 -13.6 ± 0.3
N, NE, NW 3 -21.0 ± 1.2
all types (1) 25 -13.6
The recent ice core δ18O level in terms of mean values of the shallow cores KCA, KCC, KCD, KCE, KCF and KCG is –12.1 ± 0.6 ‰. This is well above the δ18O summer level from fresh snow samples despite the cores are expected to contain 10-20% of isotopically light winter snow. The recent ice core δ18O level is also well above the expected δ18O level of –16.2 ‰ calculated from the estimated seasonal cycle of stable water isotopes according to 2.4 and the estimated seasonal distribution of precipitation in ice cores according to Figure 6.5. This enhanced recent ice core δ18O levels could be caused by two reasons:
1) Post depositional effects Despite the low molecular solid diffusion rates, the isotopic composition of snow can basically be changed by post depositional effects. Moser and Stichler (1974) reported that evaporation of snow at day time rises the δ18O levels and decreases the deuterium excess, while the condensation of air moisture at night tends to compensate this effect. Thus basically the isotope history of advected air moisture has an imprint on the isotopic composition of a snow cover. However, at CG stake measurements indicated that during 6 weeks in mid- summer 1998 the original snow height was insignificantly changed at Colle Gnifetti. This suggests that the evaporation flux is small. Moreover, the D-excess of fresh snow samples perfectly matches the 20th century mean of the deep cores. This suggests that the enhanced recent ice core δ18O level can not be explained by post depositional effects.
2) Selective conservation of precipitation events at different weather types. Figure 6.3 points out that at the low accumulation sites even during the summer months only a minor part of precipitation is conserved. Thus the enhanced δ18O ice core means could be explained if the weather types associated with high δ18O levels, L and SW are 68 6. Sampling trait
overrepresented. Specifically for the convective events L this would be very plausible as they occur at temperatures close to 0°C and they are thus expected to undergo rapid sintering. As a consequence, compared to the distribution in annual precipitation according to Figure 6.9, the Monte Rosa low accumulation cores contain a higher percentage of snow from advection of air masses from south-west and from convective events.
6.3 Summary
The seasonal distribution of precipitation in ice cores from CG was assessed mainly from the empirically estimated seasonal variation of melt events and the mean seasonal variations of precipitation amount in the Monte Rosa massif. For the comparison to instrumental records in chapter 7 the distribution according to e = 0.2 (Figure 6.5) is accepted as a best guess. This guess is based on annual accumulation and winter snow content for KCH, CC and KCS. The actual seasonal distribution of precipitation amount in ice cores may deviate from the estimation due to several reasons. As the slow sintering of firn and the formation of ice crusts without melt events was not taken into account, the seasonal distribution of precipitation may be less strongly concentrated on the summer months. Due to the influence of direct solar radiation the seasonal distribution is expected to depend on the aspect of the drill site (Figure 6.2). As for the annual accumulation, a considerable scattering in the monthly apportions for individual years is to be expected. A large part of the finally conserved precipitation originates from the advection from south- west. 69
7. Stable isotope long term records
7.1 Dated sequences
Figure 7.1 displays the δ18O depth profiles of the down to bedrock cores KCH, CC and KCS. The most striking feature are the low δ18O values in the bottom layer of ~ 5 m thickness. Based on these depth profiles, time series were established according to the dating described in chapter 5. The records KCH and KCS, ending in 1993 and 1994 respectively, were connected to the surface of the year 2000 A.D. using the shallow cores KCC and KCG, respectively (compare Figure 7.5). The CC time series ending in 1982 had already been prolonged till 1994 by the shallow core CCB. The full range of the resultant time series is shown in Figure 7.2.
A qualitative comparison of the three time series shows that major trends agree well back until 1750. However, the KCH record between 1740 (at a depth of ≈ 35 m w.e.) and the last dust horizon matched to the KCS record (1627) exhibits generally high δ18O values which do not occur in the CC and KCS time series. Moreover, below this dust but still within the range where an age constraint from annual layer counting exists (≈ 40 m w.e.), KCH shows a steep decrease in δ18O values which is totally different from CC and KCS. For KCH one core section was lost at a depth of 36.32 – 37.17 m w.e. Interpolating the annual thickness of the lost core section, a time window of 1400 – 1630 is obtained for the depth of 40 m w.e. (when starting the interpolation from the Saharan dust 1901). The dating error for KCS is less than 15 years in this period. Therefore, if the age constraint from annual layer counting is true, for KCH neither the δ18O maximum below 35 m w.e. nor the δ18O minimum near 40 m w.e. can be matched to a comparable signal in the KCS cores. As shown in Figure 7.1, the δ18O minimum of KCH near 40 m w.e. is already part of the isotopically light basal layer and the enhanced stable isotope values in KCH above correspond to the generally isotopically enriched sections above.
This extremely remarkable deviations suggest to subdivide the discussion in: 1. dated sequences where at least two cores show common trends (this chapter). 2. the basal layer signals (the following chapter). 70 7. Stable isotope long term records
re la tiv e d e p th 0 .0 0 .5 1 .0 -5 c o re 16 2 7 lo s t -1 0
-1 5
-2 0 K C H -2 5 0 1 0 2 0 3 0 4 0 -5 1 96 3 1 90 1 1 7 07 80 0 ?
-1 0 m o t t o ] b
‰
-1 5 e [
r
o O c 8 1
δ -2 0 C C -2 5 0 1 0 2 0 3 0 4 0 -5 9 43
-1 0
-1 5
-2 0 K C S -2 5 0 2 0 4 0 6 0 8 0 d e p th [m w .e .]
Figure 7.1: δ18O profiles of KCH, CC and KCS. Grey lines indicate major time markers (tritium maximum 1963, Saharan dust 1901 and Vesuvius 1707) and the last dated horizon for the respective core. Below the red dotted lines the KCH time series disagrees with KCS. Dashed green line: end of age constraint from annual layer counting. The resolution is 4-10 cm in the upper core sections and 2 cm in lowermost 10 m. 71
1627
-10
t s o l
e r o c
-15
KCH -20
-10 ] ‰ [ -15 O 8 1
δ
CC -20
-10 943
-15
KCS -20 2000 1800 1600 1400 1200 1000
Figure 7.2: δ18O time series. Grey: raw data. Black: trends highlighted by robust cubic spline smoothing (p=0.05). Below the red dotted line the KCH time series deviates from the KCS record. Dashed, green: end of age constraint from annual layer counting. 72 7. Stable isotope long term records
7.2 Systematic upstream effects
7.2.2 Backward trajectories As indicated in Figure 1.3, the deep cores KCH and KCS and the CC are approximately on a surface flow line starting near Signalkuppe. This configuration results in a relatively well defined and homogeneous common source region. The cross-section along the surface flow line shown in Figure 7.3 illustrates that snow from a certain source point moves along a flowline to the cores.
4600 Signalkuppe
4550 Bergschrund KCH 4500 0.25 m w.e./yr CC ] 0.24 m w.e./yr KCS m
[ source point
. 0,52 m w.e./yr l
s 4450 a
e d
u 1963 t i t l 4400 a bedrock 1901
4350
4300 0 100 200 300 400 position along flowline [m]
Figure 7.3: Cross-section along the surface flow line. Dots at the surface indicate surface elevations from surveyed stakes. Bedrock shape between the deep cores was adapted from Lüthi (2000). Red arrow: flowline. Grey: Isochroneus layers. 73
Figure 7.4: Backward trajectories for the Colle Gnifetti cores calculated by Lüthi (2000) using a 3D ice flow model. Black points indicate the source point at the glacier surface of an ice sample 2.5m, 5.5m, 12.5m, ... above the glacier bed. The thin lines are the corresponding flow lines. The thick line in the lower right corner indicates the bergschrund. Swiss notation: B95-2 =KCS, B82-2 = CC, B95-1 = KCH.
This source points for KCH, CC and KCS were determined from backward trajectories by Lüthi (2000) using the 3D ice flow model of Colle Gnifetti (see section 4.1). Figure 7.4 shows that, according to the flow model, the core KCH (B95-1) is situated approximately 20 m to the north-east of the surface flowline through CC (B82-2) and KCS (B95-2). However, (compare Figure 4.7), only small spatial changes in annual snow accumulation occur in this area thus from this point of view it is appropriate to consider the upstream effect is treated two dimensionally.
In the lower part of the flowline the observed surface velocities are smaller and directed slightly more to the west compared to the modelled ones (Lüthi 2000). Thus it can be expected that for KCS the calculated backward trajectories are rather too long and the real source region is further toward north-east. The uppermost part of the slope at Signalkuppe is 74 7. Stable isotope long term records
separated from the bulk glacier by the bergschrund (see Figure 7.3). According to Figure 7.4, the source region of the deep cores does not include the steep slope. However, the non-steady state movement of the steep slope above the bergschrund (section 4.2) is not described by the flow model. Therefore the potential effects of this source region anomaly are separately discussed in chapter 8.
7.2.2 Shallow cores To asses the variability in isotopical and chemical composition in the expected source region of the deep cores, nine new shallow cores were drilled between 1998 and 2000 within this work. Also the already existing shallow cores (Preunkert 1994, Schäfer 1995) were included in the evaluation. The locations of the shallow cores are shown in Figure 7.5.
86600 shallow core (new) shallow core (old) deep core
KCF KCS KCG HK CCA CCB CC 86400 CCC CCE KCD KCE CCD KCH BSK KCC KCA Tunnel BSK2 Steilhang
Bergschrund Signalkuppe Surface flow line 86200 633800 634000
Figure 7.5: Positions of shallow cores.Co-ordinates are in meters and correspond to the official Swiss coordinate system. Contour lines (10 m) indicate altitude above sea level.
The new shallow cores are focused on the upper part of the surface flow line as being the source region of the deep cores. This included each one core above (“Steilhang”) and below the bergschrund (“BSK2”) in the steep slope. Three cores (“KCD”, “KCE” and “KCF”) were drilled slightly north-east of the surface flowline as the actual source region of KCS is expected to be shifted in this direction. “KCF” is at the point where the absolute minimum of snow accumulation is found (Alean et al. 1983). The cores KCC and KCG were drilled at the 75
position of KCH and KCS. Thus the long term records could be connected to the surface of the year 2000. This is important both to affirm recent trends in antropogenic aerosol species (Klein 2001) and to inspect the signal of the pronounced warming trend of the 90th ‘s in the stable isotope records. For the new cores KCA, KCC, KCD, KCG and Steilhang equal time slices could be identified by a visible dust layer originating from 1995. From these shallow cores the mean δ18O for this distinct time slice can be calculated. Unfortunatelly in the cores BSK2 and the core KCF no horizon could be matched to other cores. Due to contamination and the extremely low annual accumulation the dating by annual counting was also not possible. From these not properly dated cores overall δ18O means were calculated. Figure 7.6 displays the spatial variations of δ18O means along the surface flow line. The Figure includes all cores except KCD, KCE, KCF and the tunnel. δ18O means are generally increasing values in the upper part of the flowline. The very large scattering stresses the influence of temporally variable snow accumulation on δ18O short term means. Therefore a mean spatial trend can hardly be derived.
-10 ’77 - ’95 ’77 - ’91 -11 ’95 - ’00 ’89 - ’93 core mean
] -12 ‰ [
O 8 1
δ -13
-14
-15 0 100 200 300 400 Position along flowline [m]
Figure 7.6: Spatial variability in δ18O along surface flow line. Equal time slices are connected by a dashed line 76 7. Stable isotope long term records
-8 ’83 - ’91 ’77 - ’95 -10 ’95 - ’00 ’98 - ’93 mean of core -12 ] ‰
[ -14 1 O 8
1 1
δ -16
-18
-20 0.0 0.2 0.4 0.6 0.8 v or A [m w.e. /yr] sub
Figure 7.7: Mean δ18O over submergence velocity. When no submergence velocity was available the accumulation rate was used.
Figure 7.7 displays δ18O means over the submergence velocity at the position of the core. The Figure suggests that δ18O means show no systematic change at submergence velocities below around 0.25 m w.e./yr. This is expected since, as outlined in chapter 5, the change of seasonal composition of ice cores with annual accumulation is weak at a low annual accumulation. At higher submergence velocities the δ18O gradually decreases due to the increasing winter snow content. The eye-fit displayed in Figure 7.7 accounts for both features.
7.2.3 Result The upstream effect in the three deep cores is now estimated in three steps: 1. The source point of each sample was determined by backward trajectories. 2. The interpolated measured submergence velocity at the source point was evaluated (see Figure 4.7) 3. The relation between submergence velocity and systematic δ18O variations as shown in Figure 7.7 was used to correct the shift in δ18O with respect to the deviation from the constant range below 0.25 m w.e./yr.
As the source region of KCH and CC is entirely in the low accumulation area of the saddle, the upstream corrected trends are virtually indistinguishable from the original ones. Figure 7.8 shows the effect of the upstream correction on the KCS time series. The upstream correction increases the overall 20th century trend from 0.5 ‰/100yr to 1.5 ‰/100yr (trends obtained by linear regression). However, the correction is negligible for the time series before 1900. For the specific source area of KCS this is mainly caused by the low spatial variations of δ18O in the upper part of the surface flow line. Additionally the change of source point with 77
age decreases for the lower core sections since the annual layer thickness decreases with depth. Therefore upstream effects should generally tend to become weaker with increasing age in alpine ice core records.
-10 ] ‰ [
O 8
1 -15
δ
uncorrected trends upstream corrected trends raw data -20 2000 1900 1800 year A.D.
Figure 7.8: Original (blue, dotted) and upstream-corrected (red, solid) trends in KCS δ18O time series. Trends are highlighted by robust cubic spline smoothing. Trends in CC and KCH are unchanged.
7.2.4 Summary and discussion The correction of upstream effects according to this procedure increases the 20th century trend in KCS more than three-fold, while long term trends before 1900 are hardly changed. Long term trends in the cores CC and KCH, as well as decadal trends in all cores are not affected. The actual source region of the core KCS is expected to be shifted towards north-east of the surface flow line. In this region δ18O means of the cores KCD, KCE and KCF show no significant spatial trend in δ18O. Moreover the submergence velocities in the triangle Signalkuppe-KCF-KCS (Figure 7.5) are generally below 0.25 m w.e./yr. (Alean et al 1984) and thus, according to Figure 7.7 no correction in δ18O is necessary. For the correction it was supposed that δ18O means depend only on the submergence velocity. Actually the δ18O means may depend on several parameters related to aspect and position. The correction does also not include a potential long term change in surface geometry of the glacier. Moreover, the correction does also not describe a potential shift of δ18O due to the origin of ice from the steep slope. In this area non steady state conditions and 78 7. Stable isotope long term records
an extraordinary variability in snow accumulation is encountered (Figure 4.7). Potential effects on the lower core sections are discussed in connection with the basal layer signals in chapter 8.
7.3 Comparison with instrumental records
7.3.1 Site specific temperature series Within ALPCLIM a homogenised and gridded Alpine long term temperature series in monthly resolution was generated. The grid covers the entire Alps at a resolution of 1° latitude and 1° longitude. The temperatures are subdivided in low level temperatures TLOW
(below 1500 m asl) and high level temperatures THIGH (Böhm et al. 2000). For the Monte
Rosa region THIGH was derived from stations up to an altitude of 2472 m asl. This is, however, still well below CG (4500 m asl.). Ice cores record temperatures at precipitation days only. Thus it is more appropriate to compare δ18O ice core records to temperature series weighted by precipitation amount. The latter are addressed as TPREC contrary to the temperatures in the climatological sense TCLI. A site specific TPREC (for CDL, 4150 m asl) was established back until 1837 by Auer et al.
(2001) as follows: The difference between TCLI and TPREC was investigated based on temperature and precipitation series in daily resolution from 1961 – 1990 using 8 stations in the Monte Rosa massif between 2000 m asl. and 2526 m asl. It turns out that TPREC is below 18 TCLI for all seasons except winter (consequently the δ O ice core record will tend to underestimate TCLI of a dry period). The mean shift TCLI - TPREC was then assessed for each month. TPREC was calculated from an long term precipitation series in monthly resolution, taken from a low level site which exhibits a high correlation with the precipitation series of the 8 high elevation stations. Additionally TPREC was shifted to the altitude of 4150 m asl using mean seasonal variations interpolated on the high altitude.
Thus three temperature series can be compared to δ18O long term records:
• TPREC back until 1837
• THIGH at the gridpoint closest to the Monte Rosa (8 ° E, 46° N) back until 1818
• TLOW at the respective gridpoint back until 1760
As the seasonal composition of precipitation in the long term records is strongly concentrated on the growing season, the δ18O time series are to be compared with accordingly weighted temperatures. Therefore the monthly temperatures of each year were weighted according to the estimated seasonal composition of low accumulation cores as shown in Figure 6.5. This is addressed as “growing season temperatures”. Such growing season temperatures were 79
calculated from TPREC , THIGH , and TLOW . Figure 7.9 shows that the trend in all three series are very similar.
3 T PREC T HIGH T 2 LOW
1 ] C ° [
T
∆ 0
-1
-2
2000 1950 1900 1850 1800 1750
Figure 7.9: Growing season temperature variability based on TPREC , THIGH , and TLOW. Thin lines: individual years. Thick lines: trends highlighted by 10 years adjacent averaging.
The high correlation of THIGH and TLOW for the summer season was pointed out by Schöner et al. (2001). The absence of long term trends in TCLI - TPREC was discussed by Auer et al.
(2001). Therefore the growing season temperatures based on the longest record TLOW will be used for the first comparison with the δ18O records.
7.3.2 Qualitative comparison Growing season temperatures and δ18O time series over the last 250 years are displayed in Figure 7.10. The ice core dating uncertainty in this period is 3 - 15 years. Moreover, annual variations are smoothed out by diffusion. Therefore, formal annual means calculated from δ18O records exhibit only a low positive correlation with annual growing season temperatures and the comparison of trends on the time scale of a few years is not appropriate. Decadal to centennial trends in all δ18O time series were highlighted by robust cubic spline smoothing. These trends in (upstream corrected) δ18O time series exhibit an impressive coherence with the growing season temperatures. They concurrently reflect the 20th century warming trend, the increased rate of warming of the last two decades and the well known warm summers of the fourties. δ18O records and instrumental series agree in overall decreasing temperatures during the 19th century. Specifically it should be mentioned that 80 7. Stable isotope long term records
according to the homogenised instrumental growing season temperatures during the last decade of the 18th century were comparable to the recent level. This feature of the homogenised Alpine records was so far not generally accepted. However, the ice core records suggest that it is to be taken for real.
Deviations between decadal trends in δ18O and temperature can be attributes to several reasons: • The δ18O values of KCS record in the period 1830-1860 are both higher than the respective levels of the two other cores or what would be expected from the temperature record. Backward trajectories indicate that this ice stems from the most convex part of the surface flow line near the CC core (Figure 7.3). The mean values 1977 –1995 in Figure 7.6 suggest that in fact in this area the δ18O values are higher than expected from the submergence velocities. • The dating error is less than 15 years for the time range shown. It is expected to be largest for the oldest part of the KCH, since for this record main volcanic horizons, particularly Tambora 1815, could not be identified. This probably explains that the pronounced maximum at the end of the 19th century occurs 1 – 2 decades earlier at KCH. 81
2
1 ] C °
[ 0
T
∆ -1 Growing season temperature -2 -10
-15
KCH -20 -10 ] ‰ [
-15 O 8 1
δ CC -20 -10
-15
-20 KCS 2000 1900 1800
Figure 7.10: Growing season instrumental temperature variations compared with upstream corrected δ18O ice core records. Grey: Raw data. Red, solid: robust cubic spline smoothing (p=0.05).
• The ice core records show a striking absolute minimum around 1895 while this signal does not occur in the instrumental temperatures. However, the occurrence in all three δ18O series suggests an atmospheric origin. The signal may be explained by a irregular seasonal distribution of precipitation amounts. Precipitation amount for Switzerland in seasonal resolution back until 1530 was estimated by Pfister (1992) on the base of documentary evidence. In fact for the period from 1885 A.D. to 1905 A.D. these series indicates a low 82 7. Stable isotope long term records
amount of precipitation in summer while spring was wet. Thus a high apportion of isotopically light snow from spring is to be expected. Temperature signal in stable isotope ice core records are generally superimposed by variation in seasonal distribution of precipitation amount (Charles et al. 1995). However, the close similarity of growing season temperatures and δ18O series for the CG cores suggests that temperature usually controls δ18O trends.
7.2.4 Sensitivity of the stable isotope thermometer Several attempts were made to quantify the relation between δ18O and instrumental temperature trends by multichannel singular spectrum analysis (M-SSA). M-SSA extracts common coherent trends from discrete scalar time series, with the only parameter is the embedding dimension M. The methodology of M-SSA is reviewed e.g. at (Plaut and Vautard 1994). It is a generalisation of the one dimensional singular spectrum analyses (SSA), which is used for trend analysis. M-SSA was applied on different combinations of time series consisting of formal annual means calculated from δ18O time series and growing season temperatures. In fact an advantage of M-SSA is that for short embedding dimensions the correlation of extracted common trends converges to 1. However, the thus obtained slopes ∆δ 18O/∆T are very critically affected by the selected time range and, even worse, depending on the selected time range the common trends are significant or not. Probably these problems are partly caused by the dating error. Anyway, the application of M-SSA was found to be not suitable and the calculated sensitivities are window dressing.
To present a best guess of the sensitivity, only the records after 1900 A.D are compared. This period exhibits a signifivant long terrm trend and a close similarity of instrumental and isotope temperatures. 20th century growing season temperatures and stacked formal annual means from the three δ 18O time series were equally smoothed by a robust cubic spline (Figure 7.11). The splines were evaluated on annual grid points. These two gridded series show a correlation of 0.9. Linear regression yields a ∆δ 18O/∆T -relations of 1.7 ‰/°C. While the value ∆δ 18O/∆T depends on the method, the sensitivity is certainly 2 – 3 times higher than the “classical” ∆δ 18O/∆T -relations of around 0.67 ‰/°C. This enhanced sensitivity is discussed below. 83
-10
-12 ] ‰ [
O
8 -14 1 δ
-16
2
1 ] C ° [
T
∆ 0
-1
2000 1980 1960 1940 1920 1900
Figure 7.11: a) Stacked formal annual means of upstream corrected δ 18O time series from KCH, CC and KCS. b) growing season temperatures (based on TPREC) . The overall 20 th century trend is indicated by robust cubic spline smoothing.
7.4 A millennium of isotope temperatures
With this estimated site specific sensitivity of the stable isotope thermometer, the temperature variability of the last millennium can be quantitatively assessed from the three δ18O long term records. These overall temperature trends are based on the δ18O records of KCH back to 1740 A.D., CC back to 1050 A.D. and KCS back to 940 A.D. Sequences below were cut as they are suspected to reflect non-climatic signals. For these time series the robust cubic spline as shown in Figure 7.10 was evaluated on a annual grid. This compensates the strongly decreasing temporal resolution of the samples in the lower core sections due to both, diffusion and annual layer thinning. Moreover, SSA for trend analysis requires an equally spaced time series. Then the mean values from this gridded data sets was calculated (indicated as mean in 84 7. Stable isotope long term records
Figure 7.12). From this mean, the first reconstructed component was calculated using SSA. This overall δ18O trends during the last millenium is shown in Figure 7.12.
individual core 1.5 -11 mean SSA 1.0 -12 0.5 ] -13 ‰ [ 0.0
O 8 1 δ -14 -0.5
-15 -1.0
-1.5 -16
2000 1800 1600 1400 1200 1000 year A.D.
Figure 7.12: 1000 years of growing season isotope temperature from stacked Monte Rosa ice core records. Grey: Robust cubic spline as shown in Figure 7.2 for the individual time series. Dotted: respective mean. Black: First reconstructed SSA component from the mean.
The temperature scale is according to a ∆δ18O/∆T -relation of 1.7 ‰/°C. The mean δ 18O for the period 1961 – 1990 was adjusted to a temperature deviation of 0°C (“recent mean”). This isotope record indicates that growing season isotope temperatures during the last three decades are at maximum level within in the last millennium. However, recent isotope temperatures are comparable to the temperature levels at the end of the 19th century, around 1700 A.D. and around 1580 A.D. According to the ice core records the Medieval warm period, which is commonly identified with the period from 1000 to 1300 A.D. exhibits an average temperature of 0.5 °C below the recent mean. The unparalleled temperature minimum is found at 1350 A.D when decadal means of isotope temperatures were more than 1°C cooler than the recent level. Other major isotope temperature minima occur at 1900 A.D and around 1615 A.D. 85
7.5 Comparison to other millennial records
7.5.1 Oscillations of Glaciers A comparison of the CG stable isotope records with extension of glaciers is tempting due to the close similarity in altitude and location. However, the extension of glaciers reflects not only temperature during ablation period but also the amount of percipitation. Moreover, the extension of glaciers generally follows climatic oscillations with a delay of several decades. The most accurate reconstruction within in the Alps are from Grosser Aletsch glacier (Bernese Alps). As the oscillations are reconstructed from the age of overridden trees, the periods of maximum extension are basically better know than the minima. Maximum extensions during last millennium (Holzhauser 1984) are indicated in Figure 7.13. These dates coincide well with main minima of isotope temperature. This may confirm that the extension of glaciers is indeed controlled by long term temperature trends. On the other hand, the coincidence suggests that major variations in the isotope temperature reflects an atmospheric signal.
7.5.2 Tree-rings Summer temperatures (April – September) are also reconstructed from tree ring density data. Thus the recorded season resembles the high elevation ice core records. The reconstructed temperatures presented here refer to the overview presented by Briffa et al. (2001). Figure 7.13 shows Northern hemisphere temperature proxies reconstructed by Jones at al. (1998), Crowley and Lowery (2000) and Briffa et al. (2001) themselves. These Northern Hemispheric temperature proxies show a close similarity to temperature reconstruction solely based on tree ring series from the Swiss Prealps (Schweingruber 1988). However, a slight difference is that the Prealp record shows high temperatures from 1660 A.D. to 1800 A.D. with a pronounced interruption around 1700 A.D. Moreover, the northern Hemispheric temperatures from Jones et al. 1999 and the frequently mentioned TLOW from Böhm 2000 is included in the Figure. These series in annual resolution were smoothed with a 50 year low pass filter. The comparison with the CG isotope temperature trends obtained by SSA is not affected by the specific smoothing procedure. Within the instrumental period, the stable isotope trends basically follow the Alpine temperatures while the tree ring series resemble the northern Hemispheric temperature. The deviations between around 1700 A.D. and 1580 A.D., when the tree rings indicate a temperature minimum but isotope temperatures remained high., may partly be explained (like for the unusual minimum in isotope temperatures around 1900) by a unusual seasonal distribution of precipitation amounts. According to Pfister et al. (1992), the decades of 1690 A.D. – 1700 A.D. and 1570 – 1600 A.D. encountered an extremely high amount of precipitation in summer. According to the seasonal variations in δ18O, this is expected to rise the stable isotope levels. Anyway, the main difference between temperature reconstruction based on the tree rings and stable isotope temperatures is in the trends before 1400 A.D. Tree rings indicate a relatively 86 7. Stable isotope long term records
constant temperature level around 0.2 °C below the recent mean. Ice core isotope temperatures exhibit a much lower level, an overall increasing trend from 1000 A.D. –1400 A.D. and a much more pronounced minimum at 1350 A.D.
CG isotope temperatures trees: Jones et al. (1998) 0.5 trees: Briffa et al. (2001) trees: Crowley and Lowery (2000) instrumental: Böhm et al. (2000) instrumental: Jones et al. (1999) 0.0 ] C ° [
T -0.5 ∆
-1.0
-1.5 2000 1800 1600 1400 1200 1000 Year A.D.
Figure 7.13: Comparison of CG isotope temperatures (black) to Northern Hemispheric temperatures reconstructed from tree ring densities and instrumental temperatures. Arrows: Maximum extension of Grosser Aletsch glacier after Holzhauser (1984).
This deviations corresponds to a shift in δ18O of around 1 ‰. Considering the recent spatial variations in δ18O suggests that these deviation may rather not be explained by a glaciological biasing. In terms of atmospheric signals, the following reasons could be possible for the low δ18O values before 1400 A.D.: • A permanently increased precipitation amount in spring. This can not be checked since the reconstruction of precipitation amount of Pfister (1992) ends in 1530 A.D. • A permanent change in circulation patterns. Specifically more precipitation events during the advection from north would lower the δ18O values (Table 6.2). At this state this also just aspeculation. • A significant difference in temperature trends between the Alpine region and the northern Hemisphere. In view of the similarity between isotope trends and Alpine instrumental temperatures after 1760, this has to be taken seriously into consideration. A constraint may arise from the evaluation of trends in chemical species. 87
7.6 Sensitivity
The ∆δ 18O/∆T –relation certainly depends somewhat on the method to calculate it. However, this does not much affect the striking feature that the mean long term ∆δ18O /∆T -relation of the CG long term records appears to be around 2.5 times higher than the classical spatial ∆δ/∆T –relation of 0.69 ‰/°C. The latter resembles both the predictions of a simple Raleigh condensation model (Daansgard 1964) and typical long term ∆δ/∆T -relations obtained for mid latitude precipitation series (Rozanski et al. 1993). Thus there is the need to understand potential reasons for the considerably enhanced sensitivity of Colle Gnifetti stable isotope thermometer, which may be adresses as follows: 1. The restriction on the growing season 2. The high elevation 3. A climate induced change in the seasonal composition of the ice cores 4. A climate induced change in precipitation types
7.6.1 The restriction on the growing season Long term ∆δ18O /∆T -relations for the Alpine region in seasonal resolution can be assessed from the 4 Swiss IAEA precipitation series. As shown in Table 7.1 a restriction on the growing season or summer months rather tends to lower the sensitivities. Thus reason 1) is ruled out. The generally low ∆δ18O /∆T –relation in summer precipitation has already been pointed out by Rozanski (1993). A low sensitivity in summer is also observed in atmospheric vapour and was mainly attributed to the continental evaporation (Agemar et al. 2001).
Table 7.1: Long term sensitivities [‰/°C] of Swiss IAEA stations in seasonal resolution. The values ware calculated by linear regression of δ 18O averages over temperature means of the respective season based on the years 1971-1992.(1): Compare Rozanski et al. (1992)
Site Year (1) Summer Midsummer (months 1-12) (months 4-9) (months 6-8)
Bern (511 m asl) 0.44 0.26 0.25 Meiringen (632 m asl) 1.03 0.73 0.58 Guttannen (1055 m asl) 0.58 0.4 0.55 Grimsel (1950 m asl) 0.81 0.54 0.56
7.6.2 The high elevation Comparison to CDD At this point it would be very tempting to compare the long term trends from CG and CDD. The high accumulation core C 10 was intensely analysed within ALPCLIM (Preunkert et al. 88 7. Stable isotope long term records
1999) and dated back to 1920. The high accumulation rate allows a proper bisection in summer and winter records. However, the all-season stable isotope long term trends are biased by the systematic change of winter snow content with depth (upstream effect). Long term trends based on the semi-annual records are biased by diffusion. The biases should be less important when focusing on strong temporal trends during shorter periods. On the other hand, the considered time period can not be too short because of depositional noise and a rest dating uncertainty. A convincing ∆δ 18O/∆T -relations could not be evaluated.
The ∆δ 18O/∆T relation from CG may be compared to the results from Fiescherhorn situated in the Bernese Alps at 4000 m asl. ((Stichler and Schotterer (2000), Schotterer et al. (1997)). At this drill site a regular seasonal distribution of precipitation is encountered. The composition of precipitation in the Bernese Alps with respect to weather type less concentrated on SW (see Figure 6.9). The reported ∆δ 18O/∆T relation is 1.1 - 1.45 ‰/°C. This suggests that the high sensitivity is rather a consequence of the high elevation than of the recorded season or weather type and that the underlying process acts on a very basic level. Therefore a closer look on the predictions of a simple Raleigh model for precipitation at high elevation alpine sites was taken. The amount of precipitation in the summit regions of the Alps being several times larger than in valleys indicates that the major part of precipitation at high alpine sites is orographic, with the main mechanism being the local forced uplift at the advection of airmasses (Böhm pers. com).
The ordinary simple Raleigh condensation model was applied on this situation. The model assumes the most simple case of vapour-liquid equilibrium condensation with immediate removal of the water from the cloud.
δ Tf, f,v, df,v δ f,p, df,p
adiabatic
H
δ isobaric δ Alps Ti, hi, i,v, di,v Ta, ha, a,v, da,v
Figure 7.14: Rainout history of an airparcel during orographic forced uplift. 89
The model was used to evaluate the isotopic composition of precipitation at high elevation sites during a situation as illustrated in Figure 7.13 An air parcel of initial temperature Ti , 18 initial relative humidity hi , and initial δ O vapour composition δi is approaching the Alps at ground level being subject to isobaric cooling until reaching the foot of the Alps at a temperature Ta. The dew point of the air parcel Td is determined by hi. In the second step the air parcel undergoes adiabatic uplift to an altitude H. Consequently the condensation starts either during isobaric cooling for Ta < Td or during the subsequent adiabatic uplift for Ta > Td .
δf and Tf denote the final isotopic composition of precipitation and temperature at the considered elevation. Figure 7.14 shows the resultant isotopic composition of precipitation δf in different altitudes if the temperature at the foot of the Alps Ta is varied and all other parameters are kept constant. It turns out that the slope ∆δf /∆Ta is much larger if the condensation starts during isobaric cooling than for a the condensation starting during adiabatic uplift.
0 0 m asl 2000 m asl -5
] 4500 m asl ‰ -10
n i
O 8 1 -15 δ [
f condensation δ starts during -20
isobaric adiabatic -25 cooling uplift
10 15 20 25 T [°C] A
Figure 7.15: δ 18O in precipitation at high alpine elevations as a function of the temperature at the foot of the Alps Ta. Initial conditions of the air parcel: Ti = 20°C, hi = 100%, δi = -10 ‰.
Table 7.2 shows the mean slopes ∆δf/∆Ta and ∆δf/∆Tf if condensation starts during adiabatic cooling. While the ∆δf/∆Ta -relation at ground level resembles the ordinaryl ∆δ /∆T relation for isobaric cooling (Daansgard 1964), an increase by almost a factor 2.5 for ∆δf/∆Ta can be observed in an altitude of 4500 m asl. However the slope in ∆δf/∆Tf is enhanced only by a factor 1.5. This indicates that one may expect a not that much enhanced sensitivity of stable isotope records with respect to a site specific TPREC. In other words, trends of TPREC at an altitude of 4500 m asl. could look more different from TLOW than for the level of 2000 m 90 7. Stable isotope long term records
asl, which was presented in Figure 7.9. The mild temperatures in the example were chosen in order to avoid all complications with the onset of vapour-solid condensation or a not well known saturated vapour pressures at extremely low temperatures. However a simple Raleigh model (Daansgard 1964) predicts that the ∆δ/∆T -relation increases at lower temperatures and for vapour-solid condensation compared to vapour-liquid condensation. Assuming e.g. a lowering of the dew point from 20°C to 0°C and vapour-solid condensation instead of vapour –liquid during isobaric cooling would increase the mean ∆δ/∆T -relation during 20°C cooling by a factor 1.7. Then ∆δf/∆Ta relations of ~ 3 ‰/°C could be quite possible.
Table 7.2: Calculated ∆δ/∆T -relations between isotopic composition of precipitation at different altitudes and temperature at the foot of the Alps and temperatures at the respective altitude during orographic forced uplift
altitude [m asl] ∆δf /∆Ta [‰/°C] ∆δf /∆Tf [‰/°C]
0 0.8 0.8 2000 1.1 1.0 4500 1.8 1.2
The strong decrease in δf at high altitudes when lowering Ta is caused by three concurrent effects: • The vapour is already more depleted during the adiabatic cooling when it reaches the foot of the Alps • The temperature difference between the foot of the Alps and high altitude becomes larger since the saturated adiabatic lapse rates (SALR) are higher at lower temperatures. Thus the fraction of condensed water vapour and consequently the isotopic depletion during the uplift increases • The higher fractionation factors at lower temperatures during the uplift additionally enhance the isotopic depletion The considered situation could be specifically important for the Monte Rosa region as a major part of the precipitation is formed during the advection of moist airmasses from south-west. Such air masses are marked by their transportation over the Mediteranean The mean surface temperature of the Mediterranean was rising by about 2°C during the last two decades. A reduced cooling and rainout of most air masses over the Mediteranean at the advection from south-west is thought to be a cause for an recent enhanced flood probability at the south slope of the Alps (Ferrara pers.com). 91
7.6.3 Climate induced changes in the ice core seasonal composition A temperature change could be correlated to seasonal variations of precipitation amount (Figure 6.4) or to the seasonal variations of the conserved fraction of precipitation events (Figure 6.3). A high summer temperature is e.g. favourable both for sintering and the formation of ice lenses. This could enhance the amount of mid-summer precipitation and thus rise the δ18O level of the conserved fraction. However, also the high accumulation core at Fiescherhorn (Bernese Alps) shows the enhanced sensitivity. This does not support the idea that it is caused by climate induced changes in the ice core seasonal composition.
7.6.4 Climate induced changes in precipitation types As shown in Table 7.3 the δ18O level in precipitation at convective weather types are distinctly higher compared the levels at advective types. Thus a temporal change in the proportions of weather and precipitation types could affect the δ18O means.
Table 7.3: Mid-summer mean δ18O weighted by precipitation amount and percentage of precipitation. The values are calculated from the fresh snow samples collected at CG.
Weather type Mean δ18O Percent of precipitation
convective –10.4 10 advective –15.8 90
Indeed, in the course of atmospheric warming during in the second half of the 20th century the annual frequency of weather types in the Alpine Region has significantly changed (Stefanicki et al. 1998). The main change is the increase of convective types during all seasons with the main increase being found for the convective high pressure types. However this weather type is generally accompanied by dry conditions and thus there are no snow samples. On the other hand the trends in the frequencies of the weather types F (convective flat) and L (convective low pressure), where there fresh snow samples were collected, are not significant in summer. Moreover, it would have to be shown that changes in annual frequency of weather types are more relevant for the high altitude.
7.6.5 Summary The comparison of the ∆δ 18O/∆T –relation of CG to Fischerhorn suggests that it is rather a matter of high elevation than of the conserved season or weather type. An enhanced sensitivity at high altitude may be partly explained from a simple Raleigh model. 92 7. Stable isotope long term records
7.7 Remarks on the D-excess
D-excess and δ 18O show a pronounced covariance throughout the time series. Accordingly, as shown in Table 7.1, the slope ∆δD /∆δ 18O within the 20th century section of all three cores is well above the slope of the meteoric water line. However, the slope in the upper core sections is not significantly different from 8. Also the slope obtained from the summer season fresh snow samples is only slightly above 8 (8.25 ± 0.1). This indicates that the enhanced slope and the resultant covariance of δ18O and D-excess is mainly a post-depositional effect due to diffusion.
Table 7.1: Slope of ∆δD/∆δ 18O in different core sections
Core ∆δD/∆δ 18O ∆δD/∆δ 18O 1900 - surface 1990 - surface
KCH 8.64 ± 0.04 8.2 ± 0.3
CC + CCB 8.50 ± 0.04 8.03 ± 0.1
KCS 8.37 ± 0.03 8.13 ± 0.14
Nevertheless, atmospheric trends of the D-excess may be extracted in two ways: 1. A modified D-excess d’ = δD – k ⋅ δ 18O could be calculated. k is then to be adapted on the respective core sections. According to Table 7.1. k depends on depth and accumulation rate. 2. The D-excess profiles could be smoothed over a period long compared to the influence of diffusion. An estimation of this distance has to consider that the diffusion via vapour phase above the firn-ice transition is much faster than the diffusion in the ice below. The strong annual layer thinning is to be taken into account as well as a potential contribution of liquid diffusion below the firn-ice transition. The main concurrent trend in all three records is a decrease of 1 ‰ in the course of the 19th century. However, the extraction of atmospheric trends from the D-excess profiles has so far nor properly been done and thus the climatic implications are not discussed.
7.8 Summary
Stable isotope records from Alpine ice cores cover at least one the last millennium while the dating has not yet been pushed to the ultimate limits. The comparison to the unique site specific instrumental temperature records shows a close coherence with an extraordinarily high sensitivity of the stable isotope thermometer. Upstream effects, as being estimated from the recent surface composition and a steady state flow model, are limited to the last century. Recent decadal means of growing season temperatures are at the warmer temperature limit within the last 1000 years. The respective range is around 1°C. The comparison of Alpine 93
isotope temperatures and northern hemispheric tree ring temperature proxies suggest that the first half of the last millennium was significantly cooler in the Alps compared with northern hemispheric temperatures. 94 8. Basal layer isotope signals
8. Basal layer isotope signals
8.1 Relation to other sites
Figure 8.1 shows a zoomed view on the δ18O basal layer signals (compare Figure 7.1).
-10 20th century average -15
6° C -20 annual -25 KCH layers
40 45
-10 m o ] -15 t t o ‰ [
b
-20 e 0 8 r
1 enhanced o
c δ -25 CC resolution
40 45
-10
-15
-20 enhanced -25 KCS resolution
65 70 75 depth [m w.e.]
Figure 8.1: δ 18O profiles in the lowermost 15% of depth range. The depth resolution was increased from 4-5 cm to 2 cm in the lower core sections.
All three down to bedrock cores from CG exhibit a bottom layer of ~ 5 m w.e. thickness characterised by δ18O values well below the respective 20th century level with an average shift of around 4 ‰. The minimum values are up to 10 ‰ below the 20th century level. The core 95
sections above shows a slightly enhanced δ18O level. In all three cores the basal layer signal is the largest signal at all and these core section is expected to represent virtually the complete conserved time range. According to the sensitivity estimated in chapter 7.3, the observed shift in δ18O of 10 ‰ corresponds to a drop in air temperature of 6°C. As this is well below Holocene long term temperature variability, it seems to be obvious to associate basal layer signals with remnants of Pleistocene ice and the isotopically heavier layer above with the Holocene climate optimum. This basic feature of δ18O basal layer records was observed in almost every down to bedrock core from cold alpine glaciers in Europe, South America, Asia and arctic islands ice caps. Without exception the signal was interpreted in terms of climate change. In fact, 14C ages of 24.000 years were measured in near bedrock sections (Thompson et al. 1998). However, the age at the Pleistocene-Holocene transition was not directly dated. These up to now results are shown in Table 8.1.
Table 8.1: Sites where basal layers depleted in heavy stable isotopes occurred.
Site Reference
Chli Titlis, Alps Lorrain and Haeberli 1990 Huascaran, Andes, Peru Thompson et al. 1995 Sajama mountain, Andes, Bolivia Thompson et al 1998
Dunde ice Cap, Qinghai-Tibetan Plateau, China Thompson et al. 1989 Guliya Ice Cap, Qinghai-Tibetan Plateau, China Thompson et al. 1997
Penny Ice cap, Baffin Island, Canada Fisher et al. 1998 Barnes Ice Cap, Baffin Island, Canada Hooke 1976 Vavilov ice cap, Severnaya Zemlya, Russia Stievenard et al. 1996 Renland ice cap, East Greenland Hansson, 1994 Agassiz ice cap Koerner and Fisher 1990 Devon Island ice cap, Canada Paterson et al. 1977
Note that this Table and the discussion in this chapter is about near bedrock signals with a typical extension of 10 m. Pleistocene records from vast polar ice shields, where the extension of the Pleistocene ice is typically 1 km, are not within this category.
It was obvious that the interpretation as Pleistocene record was also the first working hypothesis for the basal layer at CG. However, as shown in chapter 7.1, the trends in stable isotope records seemed to disagree in the lower parts of the dated sequences. Thus there is the need to discuss alternative causes for stable isotope signals in the lower core sections and two other reasons may be possible. The three potential causes are: 96 8. Basal layer isotope signals
1. Pleistocene ice 2. Source region anomaly 3. Migration of isotopes Theses hypothesises are discussed for the CG basal layer signals.
8.2 Pleistocene ice
8.2.1 Trace gases in air bubbles Trace gas measurements performed at air trapped in the bubbles of glacier ice are potential 13 15 age constraints. The chemical (CO2, N2O, CH4, δAr/N2, δO2/N2) and isotopical (δ C, δ N δ18O) composition of the trapped air in samples from KCH and KCS has been analyses by Dällenbach (2000) and Lang (2000). Special interest as dating tools is given to methane and δ18O (air). Atmospheric methane concentrations exhibit large global variations during the Holocene- Pleistocene transition. Thus basically the methane records from mid latitude ice cores could be matched with the respective well dated records from Greenland ice cores. However, it turned out that methane exhibits an increase above atmospheric levels in the basal layer section of KCS (Dällenbach 2000). Thus methane was not applicable as a dating tool for the basal layer sections. δ18O (air) reflects the global ice volume. Thus, like for methane, Alpine and Greenland ice core records could basically be matched in order to obtain age constraints for the Alpine records. One sample at a depth of 44.7 m w.e from the basal layer section of KCH was analysed. According to Figure 8.1 this is right at a section where δ18O (ice) is well below the Holocene level. A δ18O value of -0.12 ‰ was found. Corresponding ages from Greenland records are 1200 – 2000 BP, 3500 – 5000 BP, or around 9000 BP. Thus a Pleistocene age is formally excluded (Lang 2000). 97
Figure 8.2 (Lang and Daellenbach, PIUB, pers. com): Isotopic composition of trace gases in the lowermost core sections of KCS.
Figure 8.2 displays an overview of the trace gase isotopic composition in the lowermost core sections of KCS. The main signals in δ18O (air) are up to +0.25 ‰ between 95 and 96 m, down to –0.4 ‰ between 96.5 and 99 m and +0.4 ‰ at 99.7 m. The positive values correspond to ages of 10000 – 11000 BP, the negative ones to 4000 – 9000 BP (Lang 2000). Typical Pleistocene values (+1 - +1.2 ‰) were not observed. However, even δ18O (air) may be subject to artefacts. According to Figure 8.2 below 95 m
KCS shows a decrease in δ(O2/N2) and total gas content accompanied by CH4, N2O and CO2 concentrations far above the atmospheric range (Dällenbach 2000). Also a strong increase in
DOC was found (Greilich 1999). The increase in CO2 being of the same order of magnitude as the decrease in O2 suggests that the CO2 is produced by chemical reactions or biological 13 activity within the ice. A biological origin of the CO2 is supported both by its δ C value and direct spectroscopic evidence for biological activity in near bedrock ice samples (Psenner 98 8. Basal layer isotope signals
pers. com.). Probably a chemical or biological process would preferably consume the light oxygen isotopes and δ18O (air) would increase. Thus the mentioned values of δ18O (air) would be too high. Therefore without any potential artefacts the difference to Pleistocene values (+1 - +1.2 ‰) is expected to be even larger.
8.2.2 Other potential age constraints The age of the basal ice can not be predicted by the existing 3D flow model. The bedrock topography is not know in sufficient detail to model ice flow disturbances caused by bedrock undulations (Lüthi 2000). These could cause folds or ice trapped in overdeepenings. For KCS it is certain that the basal layer does not consist of stagnant ice, as the bore hole displays a strong tilt rate in the basal layer section (Lüthi 2000). Moreover, the analysis of the ice structure indicates that the ice is active (Montagnat and Duval pers. com.) A dating attempt via 14C in organic carbon contained in the ice matrix is still in progress. Thus so far the only age constraint for the basal layer sections are the findings from δ18O (air). These results do not support a mid Pleistocene age.
8.3 Source region anomaly
Figure 8.3 shows a zoomed view on the origin of the surface flow line at the north-west slope of Signalkuppe (compare Figure 1.3). In this area the bergschrund separates the moving ice at the head of a glacier from the step slope above. This steep slope moves jerky as discussed in chapter 4.2. Thus, as illustrated in Figure 8.4, the lower part of the cores could contain a mass contribution of precipitation dumped into the bergschrund or of firn originating from the steep slope. The extension of the layer originating from the source region anomaly was estimated from the expected mass flow rates. Special field activities were devoted to the isotopic composition of bergschrund and step slope. 99
86400 shallow core (new) KCH shallow core (old) KCC KCE deep cores
BSK
Aperture
KCA Tunnel BSK2
86300 Steilhang Bergschrund
Surface flow line
Signalkuppe
633900 634000
Figure 8.3: Zoomed view on the origin of the surface flow line. The foot of the step slope is between KCA and BSK2. Contour lines indicate a height difference of 10 m.
recent firn of the steep slope
F2
F3
F1 KCH CC
KCS
h2
Figure 8.4: Sketch of the cross section along the surface flow line (compare Figure 7.3) 100 8. Basal layer isotope signals
8.3.1 Mass flow rates
The mean mass flow rates per unit width from the steep slope F1 (see Figure 8.4) can be estimated in three ways:
1. F1 = inventory / mean residence time. The latter may be estimated to be less than several decades from the fact that a newspaper was found at the bedrock of the tunnel 1989 (Figure 8.3).
2. F1 = annual accumulation of steep slope times horizontal extension. Annual accumulation was taken from the Steilhang core and stake readings.
3. F1 = Thickness of steep slope times horizontal ice velocity below the bergschrund Concurrently the three methods yield a mass flow rate per unit width of 2.5 – 5 m2 w.e.a-1. For each core the mass flow rate per unit width originating from the bulk glacier surface F3 may be estimated as F3 = mean accumulation rate times distance to the bergschrund. This estimations result in a mass flow originating from the steep slope at the drill sites of 5 – 10
%. As the speed of the ice increases with height above the bedrock, the height h2 (Figure 8.4) is expected to be a higher percentage of the respective depth. Assuming that the observed one precipitation event dumped into the bergschrund during 30 years is long term representative (see below) and estimating the amount of snow per unit width in the bergschrund to be about 1 m2 w.e., the mass flow rate from the bergschrund is 1 – 2 orders of magnitude lower than that from the steep slope. According to these simple estimations, the steep slope may indeed be the source region of the basal layer. The contribution of the bergschrund is probably negligible.
8.3.2 Isotopic composition A potential contribution of crevasse fillings on the stable isotope variations measured in the glacier downstream has been discussed by Sharp et al. (1960). This seemed to be of minor importance for the CG records since the bergschrund at the Signalkuppe was never found to be substantially open since 1977 (Wagenbach pers. com.). However, in September 1999 apertures (width typically 1 m) at the glacier surface were noticed. The position is indicated in Figure 8.3. Being aware of a potential effect of the bergschrund, a close inspection was performed. Surprisingly, a descent down to the bedrock at a depth of ~18 m was possible without any digging. As indicated in Figure 8.5 (left) in the upper part of the crevasse several snow bridges were found. The lower part was a cavity ~ 5 m wide with the rock face of Signalkuppe being exposed at the slope side (Figure 8.5 right). This simple access to bedrock might be important for future basal layer sampling. Several kilograms of condensed water vapour were found mainly at the ceiling of the cavity. The relatively bright snow blown into the apertures of the bergschrund can be clearly distinguished from the condensate and the darker firn of the bulk glacier. Samples were taken from the snow bridges in different depths and from the condensate. Their stable isotope level is listed in Table 8.2. 101
Snow bridges Signalkuppe
Bulk glacier Steep slope
Condensate
Cavity
Bedrock
10 m
Figure 8.5: Left: Sketch of the bergschrund. Right: Photo from the section indicated in the sketch shows the view inside the lower part of bergschrund at the Signalkuppe. Bright: Snow blown into the Bergschrund.
Specifically devoted to the composition of the steep slope, the shallow core “BSK2” has been drilled right below bergschrund in 1998 as indicated in Figure 8.3. Additionally a snowpit “Steilhang” was dug in 2000 well above the bergschrund. Moreover, a tunnel through the steep slope was dug in 1989, which reached the bedrock at a depth of 5 m. All these samples represent a multi-annual mean. Table 8.2 shows the respective δ18O means compared to the δ18O means of the deep cores. According to Table 8.2 the isotopic composition of the steep slope rather resembles the recent core levels than the basal layer. As the samples on the surface flow line show a slightly enhanced δ18O value the steep slope may contribute to the isotopically enriched section of the layer. The mean δ18O of snow blown into bergschrund is well above the basal layer minima. However, this finding may not be representative for longer periods. Anyway case the respective mass flow rate is too small to explain the basal layer anomaly. 102 8. Basal layer isotope signals
Table 8.2: δ18O values from steep slope and bergschrund compared to the δ18O means of the deep cores.1:Includes only firn samples. A thin layer of ice covering the bedrock tends to exhibits slightly higher δ18O values.
Sample δ18O [‰]
BSK2 -12.0 Steilhang -12.41 Tunnel 1989 (1) -14.4 Snow from bergschrund -14.8 deep cores: 20th century mean around -13 deep cores: Basal layer around -17
Thus neither steep slope nor bergschrund contribution can explain the basal layer stable isotope signals at CG. Also the basal layer phenomenon in general as listed in Table 8.1 does not seem to be connected to a specific topography of the sampling site.
8.4 Migration of isotopes
8.4.1 Heuristic preliminary remark Figure 8.6 shows again the KCS basal layer δ18O record. The “deficit” of heavy isotopes (in terms of area between the graph of δ18O in the lowermost ≈ 9 m w.e. of ice and 20th century average) is apparently much larger than the “surplus” in the layer above. However, when considering the isotopic composition of the ice flow (passing through a vertical plane at the core position), deficit and surplus are to be weighted by the horizontal ice velocities. As the horizontal ice velocity increases with distance from bedrock, it may be suspected that the isotopic composition of the ice flow still matches the 20th century average. Just that is expected if the observed signal is caused by a hypothetical flow of heavy isotopes out of the isotopically depleted basal layer into the enriched layer above. 103
ice speed increases -10
surplus 20th century level ] k ‰ c [
o
deficit
-15 r 0 d 8 1 e δ b
-20 70 75 80 depth [m w.e.]
Figure 8.6: Alternative view of KCS basal layer stable isotope record
It is very simple to propose that the basal layer signals are caused by the migration of isotopes. However, it is much harder to propose a convincing underlying process which is: • effective enough to explain the magnitude of the signals • supported by other data and not in obvious contradiction to the properties of cold ice The suggestion presented here was motivated by the following items:
8.4.2 Stratigraphy and density As ice layers or Saharan dust layers are no longer visible or detectable by digital image processing, the stratigraphy of the lower third of the ice cores is rather homogeneous (Scholze 1998). However, in the lowermost ≈ 7 m w.e. of KCH and KCS a considerable variability in geometry and density of air bubbles is observed. Both KCH and KCS exhibit distinct layers of numerous large and elongated bubbles. These bubbles have a length of several mm and a diameter 1-2 mm whereas the typical bubbles in the core sections above are spherically with a diameter smaller than 1 mm. The layers of “elongated bubbles” are indicated with grey lines in Figure 8.7. On the other hand, sections almost free from bubbles are also remarkable. The last visible ice lens being several tens of meters above suggests that the bubble free layers can not be a remnant of surface melting. Both KCH and KCS exhibit discrete layers of elongated bubbles with the principal axis being inclined with respect to the core axis (“inclined bubbles”). The main horizons of these “inclined bubbles” are indicated in Figure 8.7. As visible in Figure 8.7, for KCS core the occurrence of inclined layers is associated with the upper limit of generally isotopically depleted layers. Also for KCH the inclined bubbles are associated with a strong signal in stable isotopes. 104 8. Basal layer isotope signals
KCS exhibits also larger cavities: Two horizontal cracks of several mm2 cross section and one till now completely closed cavity of several cm3 volume were found in the lowermost 1.74 m. However, since the such larger cavities were not seen in the KCH core it is assumed that they are not directly related to other basal anomalies.
-10 core average -15
-20
-25 KCH
40 45 zoomed view: Figure 8.8 -10 m ] -15 o t t ‰ o [
b
-20 0 e 8 r 1
enhanced o δ -25 CC resolution c
40 45
1 2 3 4 5 6 -10
-15
-20 enhanced -25 KCS resolution
65 70 75 depth [m w.e.]
Figure 8.7: Basal layer stable isotope signals compared to major stratigraphic events. No such stratigraphy is available for CC. Numbers at the KCS-profile: samples for ice structure. Yellow: inclined bubbles. Grey: elongated bubbles
Figure 8.8 shows a zoomed view on the stratigraphy of KCH with the high resolution density profile. Three sections of ≈ 1 cm thickness exhibit a pronounced drop in density, reaching values as low as 0.1 g/cm3 (!). It is tempting to consider this just as an error of measurement (e.g. occurring at the transition between sequencing core sections). However, the comparison 105
to the findings of visual stratigraphy indicates that these values are to be taken for true, since exactly at these sections the numerous elongated bubbles are found. While at first glance these “elongated bubbles” may resemble ordinary air bubbles, the extremely low density strongly suggests that that they were not formed as usual during the firn-ice transition.
almost free from bubbles
inclined bubbles
elongated bubbles
-10 1.0
-15 ] 3 m ] c / g ‰ [ [
0.5 y O t 8 i 1
s δ -20 n e d
-25 0.0 57 58 depth [m]
Figure 8.8: Comparison of stable isotopes, high resolution density profile and visual stratigraphy in a basal layer section of the KCH core.
Figure 8.8 shows that the layers of elongated bubbles and low density are associated with low isotopes, while the layers with few bubbles are associated with enhanced isotope values. The stratigraphic features as per description may as well be connected to ice deformation. The flow law of ice (eq. 4-1) does not include the feedback of ice deformation on ice structure, which again affects the flow parameter A. In fact, after the ice was subject to either simple shear or pure shear for a long time it becomes softer with respect to deformation by simple shear (Nye 1960). Due to this positive feedback, the gradients in horizontal ice velocities tend to concentrate on discrete shear planes. Accordingly the profile of the horizontal velocity as qualitatively illustrated in Figure 3.4 is to be considered as a smoothed approximation. Such concentration of creep in very narrow bands may even lead to the rupture of the material accompanied by open fractures and the existence of stagnant ice below the shear band (Colbeck et al. 1978). Generally fully ductile creep of polycrystalline ice without cracking is only observed at strain rates (or stresses, respectively) below a certain threshold value. At 106 8. Basal layer isotope signals
higher deformation rates brittle failure occurs (Petrenko and Withworth 1999). The investigation of the formation of cracks in laboratory experiments is summarised by Hobbs (1974). These experiments show that the formation of cracks starts at pressures of around 106 Pa, which is the range of basal stresses expected at CG. Specifically Gold (1963) noted the formation of small cavities at grain boundaries and triple junction. It may be suspected that the observations of Gold correspond to the “bubble-like cavities” as described above. To summarise, it is suspected that the observed bubble-like cavities are generated by shear planes and that these layers correspond to low isotope values.
8.4.3 Ice structure and fabric Ice structure and fabric (distribution of the c-axes and grain size) was analysed for 6 near bedrock samples from the KCS core at the depths indicated in Figure 8.7. Sample 1 and 2 indicate that the ice was subject to pure shear whereas sample 3 – 6 indicate simple shear. Besides sample 3 indicates a nearly horizontal shear plane and the samples 4 – 6 show that the ice flow is inclined at about 45° (Montagnat and Duval pers. com.). Also the tilt rates of the KCS borehole indicate that - as expected - the ice undergoes simple shear near bedrock (Lüthi 2000). Thus it can be suspected that the isotopically depleted basal layer corresponds to the zone of simple shear, while the isotopically enriched layer above is subject to pure shear.
8.4.4 The liquid phase in polycrystalline cold ice Magnetic resonance measurements indicate the presence of a liquid phase in polycrystalline ice at temperatures well below the freezing point (Clifford 1967). The observed liquid proportions, are e.g. 0.1 % for pure vapour deposited ice at –12°C (Paren and Walker 1971). The persistence of the liquid is mainly caused by freezing point depression at curved grain surfaces. Additionally quasi-liquid surface films and the presence of impurities contribute to the liquid content. The main part of the liquid is thought to be present in “triple junctions”; these are veins where three ice grains meet (Mader 1992). Depending on ice fabric and impurity content, the veins may form a continuous network that renders the ice slightly permeable to water. A consequence for the distribution of stable isotopes in cold ice is already known: The effective diffusion rate calculated from the smoothing of isotope profiles in Holocene Greenland ice is 1 - 2 orders of magnitude higher than the molecular diffusion rates for single crystals. This has been attributed to the contribution of liquid diffusion in triple junctions (Nye 1998). Ice in equilibrium with water is generally isotopically enriched with respect to the water. However, it is doubtful whether the fractionation factors for pure ice and water at 0°C (Lehmann and Siegentaler 1991) are applicable on the solid (grain) - water (triple junctions) equilibrium in cold ice since the temperatures are well below the freezing point, the hydrostatic pressure is well above the atmospheric level, impurities are present and the surfaces are curved. There is e.g. evidence that the equilibrium freezing slope for seawater is slightly higher than for pure water (Beck and Münnich pers. com.). 107
8.4.5 Suggested process Ice structure analysis and visual stratigraphy suggest a connection of ice deformation and isotopic composition. A priori it is not clear what is reason and what is consequence. The effect of chemical ice composition on ice structure and thus indirectly on ice deformation is well known (Paterson et al. 2000). However, it may be also vice versa: strain rate gradients affect the distribution of stable isotope and impurities in cold glacier ice. With the points above in mind, one could imagine that the process works as illustrated in Figure 8.9: Cavities (mainly the elongated bubbles and inclined bubbles) are formed within the shear planes of glacier ice. These cavities are preferably filled through the triple junction with liquid originating mainly from nearby pure shear layers. As the liquid is depleted in heavy isotopes, this causes a net flow of heavy isotopes out of the shear planes into the neighbouring pure shear layers. To sustain the flow of the liquid, it is essential to assume a continuing generation of cavities in the shear planes and that the liquid in the pure shear layers is supplied by sustaining a certain equilibrium grain surface curvature at the triple junctions, increased by the deformation of the grain and reduced by melting.
Pure shear: squeezing out of liquid flow of liquid
Shear panes (simple shear): formation of cavities refilled by liquid
Figure 8.9: Illustration of the suggested process potentially accounting for the migration of heavy isotopes out of the shear planes of cold ice.
8.4.6 Predicted effect on isotope profiles The suggested process is now investigated in terms of magnitude of the signals and of potential effects on the D-excess. The relation between a potential liquid flow and the isotope content can derived from considering in- and outgoing fluxes of solid ice and liquid of a small volume of glacier ice in a fixed co-ordinate system. The overall isotope balance of the volume is:
= −∇ ⋅ − ∇ ⋅ + ⋅ ∆ R (R v) (R f ) D R (8-1) E F with: v = velocity of the solid ice f = flow rate of the liquid 108 8. Basal layer isotope signals
R = overall isotope ratio of the volume
RE = isotope ratio of the ice matrix
RE = isotope ratio of the liquid phase
The first two terms account for the advection of solid ice and liquid, respectively. The last term reflects diffusion with D being an effective diffusion constant for the solid-liquid compound system according to Nye (1998). An essential part of the problem is the time constant τ of isotope exchange between veins and grains by diffusion. This time constant can be estimated from the molecular solid diffusion constant D ≈ 3 ⋅ 10-11 cm2/s (Jean-Baptiste et al. 1998) and the typical grain radius in the basal ice of g ≈ 3 mm (Montagnat and Duval pers. com.): g 2 τ ≈ ≈ 10years 4 ⋅π 2 ⋅ D However, to illustrate some aspects of the problem the simplification the instant isotopic equilibrium of solid and liquid is assumed: = α ⋅ RE RF (8-2) Additionally is assumed that the cavities are immediately filled by liquid and a unit density
for solid and liquid:
div f = −div v (8-3)
Using (8-2) and (8-3), (8-1) can be written as:
R 1 R = −v ⋅ ∇R − f ⋅∇( E ) + ∇f ⋅ R ⋅ (1− ) + D ⋅ ∆R (8-4) E α E α or in terms of delta notation:
δ
δ = − ⋅∇δ − ⋅∇ E + ∇ ⋅ δ + ⋅ − 1 + ⋅ ∆δ v f ( ) f ( 1000) (1 ) D (8-5) E α E α
The first term reflects the advection of ice while the second accounts for the advection of liquid. The generation of liquid and the refilling of cavities by the liquid phase is contained in the third term. The last term is the contribution from diffusion. Eq. (8-5) will be used below to estimate the magnitude of signals caused by migration. a) Magnitude of the signals. The amount of liquid flow and thus the efficiency of the proposed process is constrained by the mechanical properties and structure of the ice matrix. Therefore the maximum possible liquid flow is estimated from two constraints: 109
Constraint 1: The maximum of liquid flow fmax is determined by gradients in hydrostatic pressure and triple junction geometry. Assuming straight triple junctions directed parallel to the liquid flow, fmax is given by: = ρ ⋅ fmax t ft (8-6) with: ρt = number of triple junctions per unit area
ft = flow in each individual triple junction
The number of triple junctions per unit area ρt may be expressed as: ρ = 1 t (8-7) A g with Ag being the average area of a grain.
Assuming laminous flow in a triple junction of circular cross section ft can be estimated as: π ⋅ (∆p) f = ⋅ R4 (8-8) t 8⋅η ⋅l with: R : an effective radius of the veins. This is expected to be around 10 µm (Mader 1992) η : viscosity of water. Order of magnitude: 0.002 N⋅s⋅m-2 at 0°C l: distance over which the flow acts ∆p: difference in hydrostatic pressure between source region of the liquid and cavities. ∆p is limited by the hydrostatic pressure in the basal layer, which is around 106 Pa.
Combining (8-6) – (8-8) yields to: π ⋅ (∆p) f = ⋅ R4 (8-9) max ⋅ ⋅η ⋅ Ag 8 l Constraint 2: Since a pure shear volume can not be the net source of liquid flow unless it is compressed, div f is constrained by the strain rates of the solid ice matrix. The flow rate of the liquid is related to the ice deformation by: = − ε div f ∑ i (8-10) i ε with i being the diagonal elements of the strain rate tensor. Thus div f is limited by the strain rates near bedrock. According to Paterson (1994) they are -1 usually in the range of 0.001 yr . Approximately the same value for εz near bedrock at CG can be estimated from an annual layer thickness of around 1 cm in core sections around 10 m above bedrock (see Figure 5.1). The magnitude of the isotope signal is now calculated for a realistic example. Assume a glacier of thickness h = 100 m and the vertical ice velocities to be: 110 8. Basal layer isotope signals
z v = v ⋅ (1− ) (8-11) 0 h The liquid flux is assumed to have the form: z−z −( 0 )2 f = k ⋅ e l (8-12) with a maximum at a depth z0 = 95 m. This assumes that the liquid moves about a distance l. Then constraint 1 (8-9) requires: k < 10-6 m2/s /l. Constraint 2 (8-10) requires k < l⋅ε = l ⋅ 3⋅10-11 1/s. For l being smaller than the thickness of the glacier the constraint arises from (8- 10). This means that the problem is not the permeability of the ice but to squeeze out the liquid and that the maximum flow fmax can be larger if one assumes a larger extension of the layer in which the liquid is generated. With this assumptions for v and f, (8-5) can be numerically integrated using a Runge-Kutta algorithm (http://www.mathematik.ch/). Boundary condition was δ (z=0) = -14 ‰.
Furthermore it was assumed: no diffusion, δ = δE , v0 = 0.1 m/yr, l = 2 m and horizontal concentration gradients are zero. The value of v0 = 0.1m/a corresponds to a realistic annual layer thickness of 0.5 cm at a depth of 95 m Equilibrium fractionation factors were adapted from Lehmann and Siegentaler (1991). The resultant profile is shown in Figure 8.11.
-13 ] ‰ [ -14 O 8 1 δ flow of liquid
advection of ice -15 90 95 100 depth [m]
Figure 8.11: δ18O profile calculated from a hypothetical flow of the liquid phase in cold ice.
With this assumptions, the magnitude of the signals is in the range of 1 ‰. Larger signals occur for lower ice velocities and a longer distance over which the liquid flow acts. To summarise, assuming instant isotopic equilibrium between solid and liquid, realistic ice velocities and the liquid flow acting over a distance of 1 m or more, the migration of the 111
liquid phase could cause basal layer isotope signals which have a magnitude comparable to the observed ones. b) Potential effects on the D-excess Figure 8.12 shows exemplary the basal layer D-excess profile in comparison to δ18O for the KCS core. In all three cores the isotopically depleted layer shows a D-excess roughly 2 ‰ below the 20th century average.
-10 20 20th century ]
‰ level [
0
8 -15 1
δ
-20 15 20th century ] ‰ [
level s s
-25 e c x e - D
-30 10
-35 70 75 depth [m w.e.]
Figure 8.12: δ18O and D-excess in the basal layer of KCS (Raw data and low pass filtered values). Values of D-excess below 8 are probably not properly measured.
The slope δD/δ18O is around 8.5 at the basal layer transition. This is just in the range of the slopes of the 20th century sections (Table 7.4), which are mainly a result of diffusion. For the interpretation of potential significant variations in the slope it has to be considered that the slope δD/δ18O is determined by the superposition of advection, flow of liquid and diffusion (see eq. 8-5). Considering only the flow of liquid, a simple equilibrium freezing slope of 6.5 is expected for the ice composition of CG. Basically constraints for the flow of liquid could be obtained from the variations of the slope. 112 8. Basal layer isotope signals
8.4.7 Relation to Ion content Figure 8.13 shows exemplary major ion contents in the basal layer section of KCS. The lowermost 3.16 m of the KCS core (lowermost 0.08 m for the KCH core and lowermost 1.5 m for CC) exhibit silty ice. While there seems to be no connection between the occurrence of isotopically depleted layers and the silty ice, the bedrock abrasion may have contributed to the ion signals in this section. Additionally some hundred pieces of gravel are found in the lower core sections, most of them being smaller than 1 mm but reaching diameters up to 1 cm. Although most gravel is found within the silty ice, the first gravel is found well above the silty ice transition. It is not clear whether the gravel above the silty ice originates from bedrock. They could also stem from the firn of the steep slope, where stones probably from the rocks at Signalkuppe are sometimes found. However, for the moment the sections below the first gravel are excluded from the discussion.
5
1500 1 2 3 4
1000 ] b p p [
4 500 O S
0
400 ] b p p
[ 200
l C
0
-10
] -15 ‰ [
O 8
1 -20 δ silty ice -25 first gravel
65 70 75 depth [m w.e.]
Figure 8.13: δ18O and major ions for the basal layer of KCS.
Focussing on the 5 major peaks in the sulphate profile, the peaks 1-4 are accompanied by a peak in the chloride profile. These peaks are most probably Saharan dust layers as they 113
exhibit also a peak in particle profile. Volcano horizons, characterised by a high acidity, are among the smaller sulphate peaks. Contrary to peaks 1-4, the largest sulphate peak 5 is not accompanied by peaks chloride or particle content and this peak is right at the onset of the isotopic depletion. The connection to the migration process is as follows: Measurements by scanning electron microscope equipped with a energy-disperse X-ray microanalyses indicate that sulphate is concentrated in the triple junctions, while chlorine is located in the ice crystals (Mulvaney et al. 1988). Thus it expected that the layer in which the liquid pours in shows light isotopes and enhanced sulphate level while chloride should be not much affected. This is observed at peak number 5, however it is not clear why the isotopically depleted layers below do not show generally enhanced sulphate concentrations. Besides, as the liquid flow may cause sulphate peaks and may affect the acidity, the “volcano layers” in the lower core sections are to be considered critically.
8.5 Summary and outlook
All three cores exhibit an isotopically light basal layer section. For various other sites this signal was taken as a sign of Pleistocene ice. As trace gase measurements on the basal layer at CG do not show typical pleistocene concentrations, they do not support the assumption of Pleistocene origin. However, both methane and δ18O (air) are subject to artefacts which are not understood. The investigation of recent surface composition indicates that the basal layer isotope signal can not be explained by a source region anomaly. Visual stratigraphy had suggested a connection of cavities and low isotope content. Cavities were assumed to be generated in shear planes near bedrock. If these cavities are preferably filled by the remaining fraction of liquid in cold ice, the isotope values in shear planes would become lower in the course of time. Assuming isotopic equilibrium between solid and liquid, it was shown the process may be effective enough to explain the observed isotope signals. A sulphate anomaly within the basal layer could be an evidence that in fact a migration of the liquid phase occurs. Basal layer ice core records are thought to be an important archive for mid- and low latitude past climate. Thus potential artefacts deserve further investigation in the future. Moreover, their might be a connection to the well known “cold spikes”7 in the lower part of the Greenland ice sheet and the ambiguous Eemian instabilities of the GRIP and GISP2 cores from Greenland (Groote et al. 1993). Thorsteinsson et al. (1995) pointed out the covariance of crystal size and stable isotope content of the GRIP Eemian record. While it was suggested that crystal size reflect the temperature, shear zones are also known to affect crystal size. Inclined layers in the GRIP core occur right at the depth where the GISP2 isotope profile does no
7 Isotopically light layers which could not be identified with cold periods. 114 8. Basal layer isotope signals
longer agree with GRIP (Thorsteinsson 1996). Such inclined layers are also observed at the basal layer anomaly of the CG cores. Potentially the migration of liquid phase could be experimentally proved using e.g. the access to the ice-rock transition at Jungfraujoch, Bernese Alps. Ice spiked with a tracer, preferably tritium, could be inserted in the basal ice and after a while a potential migration of the tracer could be investigated. Laboratory experiments with a glacier simulator, as shown in Annex D were not helpful to study the migration process. 115
References
Agemar, T., Zahn, A. and Sonntag, C.. An 18-year record of stable isotopes in atmospheric water vapour at Heidelberg: New implications for local deuterium excess. To be published.
Alean, J., Haeberli, W. and Schädler, B., 1983. Snow accumulation, firn temperature and solar radiation in the area of the Colle Gnifetti core drilling site (Monte Rosa, Swiss Alps): distribution patterns and interrelationships. Zeitschrift für Gletscherkunde und Glazialgeologie 19(2): 131-147.
Armbruster, M., 2000. Stratigraphic dating of high elevation alpine ice cores over the last 1000 years (in German). Master thesis. Institut für Umweltphysik. University of Heidelberg.
Armendaud, A., Genthon, C., 1994. Modelling global distributions of 222Rn, 210Pb and 10Be in the atmosphere with general circulation models. IAEA-SM-329/6: 15-23.
Auer, I., Böhm, R., Huhle, C. and Schöner, W., 2001. Precipitation adjusted temperature series - a way to optimise instrumental climate time series for comparisons with stable isotope ice core information from high Alpine sites. To be published.
Barry, R. G. (1981). Mountain weather and climate, Methuen.
Barlow, L. K., White, J. W. C., Barry, R. G., Rogers, J. C. and Grootes, P. M., 1993. The north atlantic oscillation signature in deuterium and deuterium excess signals in the Greenland Ice Sheet Project 2 ice core, 1840-1970. Geophysical Research Letters 20(24): 2901-2904.
Beck, N. (1985). Laboratory experiments on the formation of solar radiation induced melt layers in dry snow (in German). Master thesis. Institut für Umweltphysik. University of Heidelberg.
Beck, N., Wagenbach, D. and Münnich, K. O., 1988. Laboratory experiments on the formation of solar radiation induced melt layers in dry snow. Zeitschrift für Gletscherkunde und Glazialgeologie 24(1): 31-40. 116 References
Bigler, M., Wagenbach, D., Fischer, H., Kipfstuhl, J., Miller, H., Sommer, S. and Stauffer, B., submitted 2001. Sulphate Record from a Northeast Creenland Ice Core over the last 1200 Years basid on Continuous Flow Analysis.
Böhm, R., Auer, I., Brunetti, M., Maugeri, M., Nanni, T. and Schöner, W., submitted 2000. Regional temperature variability in the European Alps 1760-1998. International Journal of Climatology.
Brand, W. A., 1996. High Precision Isotope Ratio Monitoring Techniques in Mass Spectrometry. J. Mass Spectrom. 5: 387-397.
Brand, W. A., Avak, H., Seedorf, R., Hofman, D. and Conradi, T., 1996. New methods for fully automated isotope ratio determunation from hydrogen at the natural abundance level. Isotopes Environ. Health Stud. 32: 263-273.
Briffa, R. B., Osborn, T. J. Schweingruber, F. H., Harris, I. C., Jones, P. D. Shiyatov, S. G., Vaganov, E. A., 2001. Low-frequency temperaure variations from a northern tree ring density network. Journal of Geophysical Research 106(D3): 2929-2941.
Ciais, P. and Jouzel, J., 1994. Deuterium and Oxygen 18 in precipitation: Isotopic model, including mixed cloud processes. Journal of Geophysical Research 99(D8): 16793- 16803.
Clausen, H. B., Hammer, C. U., Hvidberg, C. S., Dahl-Jensen, D., Steffensen, J. P., 1997. A comparison of the volcanic records over the past 4000 years from the Greenland Ice Core Project and Dye 3 Greenland ice cores. Journal of Geophysical Research 102(C12): 26707-26723.
Clifford, J., 1967. Proton Magnetic Resonance Data on Ice. Chemical Communications: 880- 881.
Colbeck, S. C., Lawrence, W. S. and Gow, A. J., 1978. Creep rupture at depth in a cold ice sheet. Nature 275: 733.
Cole, J., Rind, D., Webb, R. S., Jouzel, J. and Healy, R., 1999. Climatic controls on interannual variability of precipitation d18O: Simulated influence of temperature, precipitation amount, and vapor source region. Journal of Geophys. Res. 104(D12): 14223-14235. 117
Cole-Dai, J., Mosley-Thompson, E., Wight, S. P., Thompson, L. G., 2000. A 4100-year record of explosive volcanism from an East Antarctic ice core. Journal of Geophysical Research 105(D19): 24431-24441.
Craig, H. and Gordon, L., 1965. Deuterium and oxygen 18 variation in the ocean and marine atmosphere. In: Stable Isotopes in Oceanographic Studies and Paleotemperatures. E. Tongiorgi. Pisa, Spoleto: 9-130.
Crowley, T. J., Lowery, T. S., 2000. How warm was the Medieval warm period? Ambio 29: 51-54.
Cuffey, K. M. and Steig, E. J., 1998. Isotopic diffusion in polar firn: implications for interpretation of seasonal climate parameters in ice-core records, with emphasis on central Greenland. Journal of Glaciology 44(147): 273 ff.
Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16: 436-468.
Dansgaard, W., Johnsen, S. J., Clausen, H. B. and Gundestrup, N., 1973. Stable Isotope Glaciology. Meddelelser om Gronland 197(2).
Demoz, B. B., Wartburton, J. A. and Stone, R. H., 1991. The influence of rimimg on the oxygen isotopic composition of ice phase precipitation. Atmospheric Research 26: 463-488.
Döscher, A., Gäggeler, H. W., Schotterer, U. and Schwikowski, M., 1995. A 130 years deposition record of sulphate, nitrate and chlorid from a high alpine glacier. Water, Air and Soil Pollution 85: 603-609.
Ehhalt, D. and Knott, K., 1965. Kinetische Isotopentrennung bei der Verdampfung von Wasser. Tellus XVII(3): 389-397.
Ehhalt, D., Knott, K., Nagel, J. F. and Vogel, J. C., 1963. Deuterium and Oxygen 18 in Rain Water. Journal of Geophysical Research 68(13): 3775 ff.
Eichler, A., Schiwikowski, M., Gäggeler, H. W., 2001. Meltwater-induced relocation of chemical species in Alpine firn. Tellus 53B: 192-203.
Fischer, H., 1997. Spacial variability in ice core time series from north-east Greenland (in German). PhD Thesis, Institut für Umweltphysik, Universität Heidelberg. 118 References
Fisher, D. A., 1991. Remarks on the deuterium excess in precipitation in cold regions. Tellus 43B: 410-407.
Fliri, F. and Schüepp, M., 1984. Synoptische Klimatographie der Alpen zwischen Mont Blanc und Hohen Tauern (Schweiz - Tirol - Oberitalien) und Alpine Witterungslagen und europäische Luftdruckverteilung. Wissenschaftliche Alpenvereinshefte, Heft 29, Innsbruck.
Gat, J. R. and Carmi, I., 1970. Evolution of the Isotopic Comosition of Atmospheric Waters in the Meditrranean Sea Area. Journal of Geophysical Research 75(15): 3039- 3048.
Gat, J. R., Bowser, C. J. and Kendall, C., 1994. The contribution of evaporation from the Great Lakes to the continental atmosphere: estimate based on stable isotope data. Geophysical Research Letters 21(7): 557-560.
Gehre, M., Hoefling, R., Kowski, P. and Strauch, G., 1996. Sample preparation device for quantitative hydrogen isotope analysis using chromium metal. Analytical Chemistry 68(24): 4414-4417.
Gold, L. W., 1963. Deformation mechanisms in ice. In: Ice and Snow, 8-27.
Greilich, S. (1998). Measuring organic carbon content of Alpine ice (in German). Master thesis, Institut für Umweltphysik. University of Heidelberg.
Grootes, P. M., Stuiver, M., Thompson, L. G. and Mosley-Thompson, E., 1989. Oxygen Isotope Changes in tropical ice, Quelccaya, Peru. Journal of Geophysical Research 94(D1): 1187-1194.
Grootes, P. M., Stuiver, M., White, J. W. C., Johnsen, S. J. and Jouzel, J., 1993. Comparision of oxygen isotope records from the GISP2 and GRIP Greenland ice cores. Nature 366: 552-554.
Haeberli, W., Schmid, W. and Wagenbach, D., 1988. On the geometry, flow and age of firn and ice at the Colle Gnifetti core drilling site (Monte Rosa, Swiss Alps). Zeitschrift für Gletscherkunde und Glazialgeologie 24(1): 1-19.
Haeberli, W., Schotterer, U., Wagenbach, D., Haeberli-Schwitter, H. and Bortenschlager, S., 1983. Accumulation characteristics on a cold, high-Alpine firn saddle from a snow-pit study on Colle Gnifetti, Monte Rosa, Swiss Alps. Journal of Glaciology 29(102): 260- 271. 119
Hammer, C. U., Clausen, H. B. and Dansgaard, W., 1980. Greenland Ice Sheet evidence of postglacial volcanism and its climatic impact. Nature 288: 230-235.
Hendricks, M. B., DePaolo, D. J. and Cohen, R. C., 2000. Space and time variation of d18O and dD in precipitation: Can paleotemperature be estimated from ice cores? Global Biochemical Cycles 14(3): 851-861.
Hennemuth B., N., I., 1986. Das Feuchtefeld in einem alpinen Endtal. Meteorologische Rundschau 39(6): 233-239.
Hobbs, P. V., 1974. Ice Physics. Oxford, Clarendon Press.
Hoffmann, G., Jouzel, J., Masson, V., 2000. Stable water isotopes in atmospheric general circulation models. Hydrological Processes 14(8): 1385-1406.
Holzhauser, P., 1984. Zur Geschichte der Aletschgletscher und des Fieschergletscher. Phys. Geogr. 13: 1-448.
Jean-Baptiste, P., Jouzel, J., Stievenard, M. and Ciais, P., 1998. Experimental determination of the diffusion rate of deuterated water vapor in ice and the application to the stable isotope smoothing of ice cores. Earth and Planetary Science Letters 158: 81-90.
Johnsen, S. J., Dansgaard, W. and White, J. W. C., 1989. The origin of Arctic precipitation under present and glacial conditions. Tellus 41B: 452-468.
Jones, P. D., Briffa, K. R., Barnett, T. P., Tett, S. F. B., 1998. High-resolution paleoclimatic records for the last millennium: Interpretation, integration and comparison with control-run temperatures. Holocene 8: 455-471.
Jones, P. D., New, M., Parker, D. E., Martin, S. and Rigor, I. G., 1999. Surface air temperature and its changes over the past 150 years. Reviews of Geophysics 37(2): 173-199.
Jouzel, J. and Koster, R. D., 1996. A reconsideration of the initial conditions used for stable water isotope models. Journal of Geophysical Research 101(D17): 22,933-22,938.
Jouzel, J. and Merlivat, L., 1984. Deuterium and Oxygen 18 in Precipitation: Modeling of the Isotopic Effects During Snow Formation. Journal of Geophysical Research 89(D7): 11749-11757. 120 References
Jouzel, J., Merlivat, L. and Lorius, C., 1982. Deuterium excess in an East Antarctic ice core suggests higher relative humidity at the oceanic surface during the last glacial maximum. Nature 299: 688-691.
Jung, W., 1993. 100 yeras of aerosol deposition on a high elevation Alpine glacier (in German).Master thesis, Institut für Umweltphysik, University of Heidelberg.
Knott, K. 1964. Seasonal variations of Deuterium and Oxygen-18-content in precipitation. Master thesis, Institut für Umweltphysik, Universtity of Heidelberg.
18 Koster, R. D., Valpine, D. P. and Jouzel, J., 1993. Continental water recycling and H O 2 concentrations. Geophysical Research Letters 20: 2215-2218.
Lal, D. and Peters, B., 1967. Cosmic ray produced radioactivity on the Earth. In: Encyclopedia of physics, ed. S. Flügge. Berlin, Springer: P. 551-612.
Lehmann, M. and Siegenthaler, U., 1991. Equilibrium oxygen- and hydrogen-isotope fractionation between ice and water. Journal of Glaciology 37(125): 23-26.
Lorrain, R. and Haeberli, W., 1990. Climatic change in a high-altitude alpine area suggested by the isotopic composition of cold basal glacier ice. Annals of Glaciology 14.
Luder, R. 1986. 10Be und 7Be in Niederschlägen und Luftfiltern, Master thesis, University of Bern.
Lüthi, M., 2000. Rheology of cold firn and dynamics of a polythermal ice stream. PhD thesis, ETH Zurich.
Mader, H. M., 1992. Observations of the water-vein system in polycrystalline ice. Journal of Glaciology 38(130): 333-347.
Majoube, M., 1971. Fractionnement en 18O entre la glace et la vapeur d'eau. J. Chim. Phys. 68(4): 625-636.
Merlivat, L. and Nief, G., 1967. Fractionnement isotopique lors des changements d'etat solide-vapeur de l'eau à des températures inférieures à 0°C. Tellus XIX(1): 122-127. 121
Merlivat, M. and Jouzel, J., 1979. Global Climatic Interpretation of the Deuterium-Oxygen 18 Relationship for Precipitation. Journal of Geophysical Research 84(C8): 5029-5033.
Meyssonnier, J., 1989. Ice flow over a bump: Experiment and numerical simulations. Journal of Glaciology 35(119): 85ff.
Moser, H. and Stichler, W., 1974. Deuterium and oxygen-18 as an index of the properties of snow covers. Snow mechanics 114: 122-135.
Mulvaney, R., Wolff, E. W. and Oates, K., 1988. Sulphuric acid at grain boundaries in Antarctic ice. Nature 331: 247-249.
Nye, J. F., 1960. The response of glaciers and ice-sheets to seasonal and climatic changes. Proc. R. Soc. A 256: 559 ff.
Nye, J. F., 1998. Diffusion of isotopes in the annual layers of ice sheets. Journal of Glaciology 44(148): 467-468.
Oeschger, H., Schotterer, U., Stauffer, B., Haeberli, W. and Röthlisberger, H., 1977. First results from alpine core drilling projects. Zeitschrift für Gletscherkunde und Glaziologie 13(1/2): 193-208.
Paren, J. G. and Walker, J. C. F., 1971. Influence of limited Solubility on the Electrical and Machanical Properties of Ice. Nature Physical Science 230.
Paterson, W. S. B. (1994). The physics of glaciers, Oxford, Pergamon Press.
Paterson, W. S. B., 2000. Why ice age ice is sometimes "soft". Cold Regions Science and Technology 20(1): 75-90.
Petit, J. R., Jouzel, J., Raynaud, D., Barkov, N. I., Barnola, J. M., Basile, I., Benders, M., Chappellaz, J., Davis, M., Delaygue, G., Delmotte, M., Kotlyakov, V. M., Legrand, L., Lipenkov, V. Y., Lorius, C., Pepin, L., Ritz, C., Saltzman, E. and Stievenard, M., 1999. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature 399: 429 ff.
Petrenko, V. F., Withworth, R. W. (1999). Physics of ice, Oxford University Press.
Plaut, G. and Vautard, R., 1994. Spells of Low-Frequency Oscillations and Weather Regimes in the Northern Hemisphere. Journal of the Atmospheric Sciences 51(2): 210-236. 122 References
Preunkert, S. (1994). Glazio-chemical properties of Colle Gnifetti compared to the regional area (in German). Master thesis, Institut für Umweltphysik. University of Heidelberg.
Preunkert, S. and Wagenbach, D., 1998. An automatic recorder for air/firn transfer studies of chemical aerosol species at remote glacier sites. Atmospheric Environment 32(23): 4021-4030.
Preunkert, S., Wagenbach, D., Legrand, L. and Vincent, C., 2000. Col du Dome (Mt Blanc Massif, French Alps) suitability for ice-core studies in relation with past atmospheric chemistry over Europe. Tellus 52B: 993-1012.
Raisbeck, G. M. and Yiou, F., 1985. 10Be in polar ice and atmospheres. Annales of Glaciology 7.
Raisbeck, G. M., Yiou, F., Fruneau, M., Loiseaux, J. M., Lieuvun, M. and Ravel, J. C., 1981. Cosmogenic 10Be/7Be as a probe of atmospheric transport process. Geophysical Research Letters 8(9).
Rogers, R. R., Yau, M.K. (1989). A short course in cloud physics. Oxford, Pergamon.
Rozanski, K., Araguas-Araguas, L. and Gonfiantini, R., 1993. Isotopic pattern in modern global precipitation. In: Climate Change in Continental Isotopic Records. P. K. Swart, K. C. Lohmann, J. McKenzie and S. Savin, American Geophysical Union: 1-36.
Rozanski, K., Araguás-Araguás, L. and Gonfiantini, R., 1992. Relation Between Long-Term Trends of Oxygen-18 Isotope Composition of Precipitation and Climate. Science 258: 981-985.
Rozanski, K., Sonntag, C. and Münnich, K. O., 1982. Factors controlling stable isotope composition in European precipitation. Tellus 34: 142-150.
Saey, P. (1998). Concentration and size distribution of mikroparticles in Alpine and polar ice cores (in German). Master thesis, Institut für Umweltphysik, University of Heidelberg.
Rozanski, K., Johnsen, S. J., Schotterer, U. and Thompson, L. G., 1997. Reconstruction of past climates from stable isotope records of palaeo-precipitation preserved in continental archives. Hydrological Sciences - Journal des Sciences Hydrologiques 42(5): 725-745. 123
Schäfer, J. (1995). Rekonstruction of bio-geochemical cycles using an Alpine ice core (in German). Master thesis, Institut für Umweltphysik. Universität Heidelberg.
Schwikowski, M., Brütsch, S., Gäggler, H. W. and Schotterer, U., 1999. A high-resolution air chemistry record from an Alpine ice core: Fiescherhorn glacier, Swiss Alps. Journal of Geophysical Research 104(D11): 13,709-13,719.
Scholze, M. (1998). Analyse der visuellen Stratigraphie alpiner Eisbohrkerne mittles digitaler Bildverarbeitung. Institut für Umweltphysik. Heidelberg, Universität Heidelberg: 79.
Schotterer, U., Haeberli, W., Good, W., Oeschger, H. and Röthlisberger, H., 1978. Datierung von kaltem Firn und Eis in einem Bohrkern vom Colle Gnifetti, Monte Rosa. Jahrbuch der Schweizerischen Naturforschenden Gesellschaft, wissenschaftlicher Teil: 48-57.
Schotterer, U., Fröhlich, K., Gäggeler, H. W., Sandjordj, S. and Stichler, W., 1997. Isotope records from mongolian and alpine ice cores as climate indicators. Climatic Change 36: 519-530.
Schotterer, U., Schwarz, P., Rajner, V., 1998. From pre-bomb levels to industrial times. In: Isotope techniques in the study of environmental change. Vienna, IAEA: 581ff.
Schweingruber, F. H., Bartholin, T., Schär, E., Briffa, K. R., 1988. Radiodensitometric- Dentroclimatological Conifer Chronologies from Lapland (Scandinavia) and the Alps (Switzerland). Boreas 17: 559-566.
Sharp, R. P., Epstein, S. and Vidziunas, I., 1960. Oxygen-Isotope Ratios in the Blue Glacier, Olympic Mountains, Washington, U.S.A. Journal of Geophysical Research 65(12): 4043-4059.
Siegenthaler, U. and Oeschger, H., 1980. Correlation of 18O in precipitation. Nature 285: 314-317.
Simkin, T. and Siebert, L. (1994). Volcanoes of the World. Tucson, Arizona, Geoscience Press in association with Smithsonian Institution, Global Volcanism Program.
Stanzick, A. 1996. Räumliche und zeitliche Depositionsvariationen der Radioisotope 10Be und 210Pb in Eisbohrkernen Zentralgrönlands. Master Thesis, Universität Heidelberg 124 References
Stanzick, A., 2001. Raum-Zeit-Variationen von 10Be, 210Pb und 36Cl in der grönländischen Firndecke: Luft-Firn Transfer und rezente Trends, PhD Thesis, University of Heidelberg.
Stefanicki, G., Talkner, P. and Weber, R. O., 1998. Frequency Changes of Weather Types in the Alpine Region since 1945. Theoretical and Applied Climatology 60: 47-61.
Stievenard, M., Nikolaev, V., Bol'shiyanov, D. Y., Flehoc, C., Jouzel, J., Klementyev, O. L. and Souchez, R., 1996. Pleistocene ice at the bottom of the Vavilov ice cap, Severnaya Zemlya, Russian Arctic. Journal of Glaciology 42(142): 403-406.
Tett, S. F. B., Stott, P. A., Allen, M. R., Ingram, W. J. and Mitchel, J. F. B., 1999. Causes of twentieth-century temperature change near the Earth's surface. Nature 299: 569-572.
Thompson, L. G., Davis, M. E., Mosley-Thompson, E., Sowers, T. A., Henderson, K. A., Zagorodnov, V. S., Lin, P.-N., Mikhalenko, V. N., Campen, R. K., Bolzan, J. F., Cole- Dai, J. and Francou, B., 1998. A 25,000-Year Tropical Climate History from Bolivian Ice Cores. Science 282: 1858-1864.
Thompson, L. G., Mosley-Thompson, E., Davis, M. E., Bolzan, J. F., Dai, J., Yao, T., Gundestrup, N., Wu, X., Klein, L. and Xie, Z., 1989. Holocene-Late Pleistocene Climatic Ice Core Records from Qinghai-Tibetan Plateau. Science 246: 474-477.
Thompson, L. G., Mosley-Thompson, E., Davis, M. E., Lin, P. N., Henderson, K. A., Cole- Dai, J., Bolzan, J. F. and Liu, K.-b., 1995. Late Glacial Stage and Holocene Tropical Ice Core Records from Huascaran, Peru. Science 269: 46-50.
Thorsteinsson, T., Kipfstuhl, J., Eicken, H., Johnsen, S. J. and Fuhrer, K., 1995. Crystal size variations in Eemian-age ice from the GRIP ice core, Central Greenland. Earth and Planetary Science Letters 131: 381-394.
Thorsteinsson, T. (1996). Textures and fabrics in the GRIP ice core, in relation to climate history and ice deformation. Berichte zur Polarforschung 205. Bremerhaven, Alfred- Wegener-Institut für Polar-und Meeresforschung.
Turekian, K. K., Graustein, W. C., Cochran, K., 1989. Lead-210 in the SEAREX program: an aerosol tracer across the pacific. Chemical Oceanography. J. P. Riley, Chester, R., Academic: 51-81. 125
Völker, G., 1996. Erfassung der natürlichen Variabilität der Spurenstoffzeitreihen in hochalpinem Firn anhand von Untersuchungen der saisonalen und räumlichen Variation von totalem Mineralstaub und 210Pb an Schneeproben des Colle Gnifetti. Master thesis, Institut für Umweltphysik, University of Heidelberg.
Wagenbach, D., 1989. Environmental Records in Alpine Glaciers. In: The Environmental Record in Glaciers and Ice Sheets. H. Oeschger and J. C. C. Langway, John Wiley & Sons Ltd.: 69-83.
Wagenbach, D., 1992. Special problems of mid-latitude glacier ice-core research. Greenhouse Gases, Isotopes and Trace Elements in Glaciers as Climatic Evidence of the Holocene, Zurich.
Wagenbach, D., 1993. Special problems of mid latitude glacier ice core research. ESF/EPC workshop: Greenhouse gases, isotopes and trace elements in glaciers as climatic evidence for Holocene time, Zürch, VAW-ETH-Zürich.
Wagenbach, D., Böhm, R., Cachier, H., Haeberli, H., Hoelzle, M., Legrand, M., Maggi, V. and Stocker, T., 1998. Environmental and Climate Records from High Elevation Alpine Glaciers (ALPCLIM). ECSC (European Commission's Science Conference), Vienna.
Wagenbach, D., Preunkert, S., Schäfer, J. and Jung, W., 1996. Northward Transport of Saharan Dust Recorded in a Deep Alpine Ice Core. In: The Impact of African Dust Across the Mediterranean. S. Guerzoni and R. Chester, Kluwer Academic Publishers: 291-300.
Wagner, S., 1994. Dreidimensionale Modellierung zweier Gletscher und Deformations- anaylse von eisreichem Permafrost. Zürich, VAW - ETH-Zürich.
Werner, R. A., Brand W.A., 2001. Referencing strategies and techniques in stable isotope ratio analysis. Rapid communications in mass spectrometry 15(7): 501-519.
Zielinski, G. A., 1995. Stratospheric loading and optical depth estimates of explosive volcanism over the last 2100 years derived from the Greenland Ice Sheet Project 2 ice core. Journal of Geophysical Research 100(D10): 20937-20955.
I
Annex A: Core meta data
Table A.1: Basic meta data of all cores which were used in this thesis. Positions were determined by survey in the course of mass balance measurements or by GPS. They correspond to the official Swiss coordinate system. (1): Estimated from sketches made during earlier field work. Accuracy is lower.
Core Drilled in Depth [m] Position y [m] Position x [m]
CC 1982 64.1 633872.0 86417.0 Tunnel 1989 5 634020.0 (1) 86330.0 (1) BSK 1991 17.7 633941.0 (1) 86364.4 (1) Hangkern 1991 10.9 633810.0 (1) 86455.0 (1) Zumsteinkern 1991 12.3 633670.0 (1) 86715.0 (1) Sattelkern 1991 6.6 633822.0 (1) 86595.0 (1) CCA 1993 20 633859.0 (1) 86434.0 (1) CCB 1993 9.3 633840.0 (1) 86432.0 (1) CCC 1993 3.9 633872.0 (1) 86400.1 (1) CCD 1993 4.1 633897.5 (1) 86374.5 (1) CCE 1993 (?) 21 633890.5 (1) 86398.2 (1) KCH 1995 60.4 633922.4 86383.0 KCS 1995 99.9 633745.7 86483.8 BSK2 1998 8.5 633970.3 86319.5 Steilhang 2000 3.6 633988.0 86298.8 KCA 2000 7.1 633960.0 86330.0 KCC 2000 4 633922.0 86381.0 KCD 2000 3.6 633944.9 86402.4 KCE 2000 1.9 633956.9 86380.8 KCF 2000 2.4 633985.0 86500.0 KCG 2000 8 633747.0 86484.0 II
Annex B: Updated volcano chronology
Table B.1: Large volcano eruptions since 876 A.D. which potentially caused volcanic fallout at CG. The collection is based on the Greenland volcano records of GISP 2 (Zielinski 1995), GRIP and Dye 3 (Clausen et al. 1997) and the historical events listed by Simkin and Siebert (1994). Years: for B and N events according to the ice core chronologies, for L events historical dates. Types: see 5.2.“N, B?” was assigned if an eruption in an Antarctic core was recorded within ± 2 years of the N event. Considered cores from Antarctica for that: Plateau remote, South Pole,Siple and Dyer, Byrd (Cole- Dai et al. 2000).
Type Year A.D. Name
N 1979 Westdhal, Alaska N 1971 Hekla, Iceland N 1969 Fernandina, Galapagos B 1966 Shevelach, Kamchatka / Agung, Indonesia N 1956 Beymianny, Kamchatka N 1924 Raikoke, Ryukyu Islands N 1917 Katla, Iceland N 1915 Sakura-jiama, Japan N 1912 Katmai, Alaska N 1908 Ksudach, Kamchatka N 1902 Santa Maria, Guatemala / Soufriere, St Vincent / Pelee, Martinique L 1875 - 1906 Vesuvius, Italy B 1883 Krakatau, Indonesia N 1854 Sheveluch, Kamtchatka / Chikurachki, Kurile Islands B 1835 Coseguina. Nicaragua N 1831 Babuyan, Philippines N 1830 Kliuchevskoi, Kamchatka B 1815 Tambora, Indonesia B 1809 ? L 1787 Etna, Italy N 1784 Laki, Iceland III
N 1766 Hekla, Iceland N 1738 Tarumai, Japan N 1731 Lanzarote, Canary Islands N 1728 Orafajokull, Iceland N, B? 1696 Komaga-Take, Japan N, B? 1694 Hekla, Iceland N, B? 1668 Tarumai, Japan N 1646 Long Island, New Guinea N, B? 1642 Awu, Indonesia (Deception, Parker?) / Komaga- Take, Japan N, B? 1639 ? N 1638 ? L 1631 - 1632 Vesuvius, Italy N 1605 Momotombo, Nicaragua B 1603 Huaynaputina, Peru N, B? 1599 Hekla, Iceland / Ruiz, Columbia N 1589 Kelut, Java N 1588 Billy Mitchel, Solomon Island N 1571 ? N 1554 ? N 1513 ? N 1509 ? N 1484 Mt. St Helens, Washington N 1482 Mt. St Helens, Washington N 1479 Veidivotn, Iceland / Mt St. Helens, Washington N 1478 ? B 1460 / 1461 Kuwae, Vanatu N 1441 ? N 1403 ? N 1360 ? N, B? 1345 Mono Craters, California N 1329 Cerro Bravo, Columbia N, B? 1286 ? B 1259 El Chichon? IV
N 1230 ? N 1228 ? N 1206 Oshima, Japan N, B? 1195 Oshima, Japan N, B? 1176 Krafla, Iceland? N 1166 ? N 1104 Hekla, Iceland N 1027 Baitoushan, China / Billy Mitchell, Solomon Islands N 943 ? N 940 Eldga, Iceland N 939 Eldga, Iceland N 937 Eldga, Iceland N 932 ? N 916 Towada, Japan N 903 Tolbachik, Kamchatka / Ksudach, Kamchatka N 901 ? N 895 ? N 889 ? N 876 ? V
Annex C: Radionuclide inventories
C.1 Background
The inventory of the radioisotopes 10Be, 210Pb, 137Cs and 3H as well defined atmospheric tracers was determined for the ALPCLIM cores from CG and CDD. These inventories basically contain information on the seasonal distribution and the source region of air masses. The relevant properties of the investigated radionuclides are listed in Table C.1 and the transport to a high elevation Alpine site is illustrated in Figure C.1.
Table C.1: Basic properties of the radionuclides 10Be, 210Pb, 137Cs and 3H.
Nuclide Half live [yr] Source Bound to Deposition
10Be 1.51 ⋅ 106 Stratosphere (Cosmogenic) Sub micron aerosol Wet and dry 210Pb 22.3 Continental troposphere Sub micron aerosol Wet and dry (Terrigenic) 137Cs 30.17 Stratosphere (Nuclear tests) Sub micron aerosol Wet and dry (larger particles)
3H 12.43 Stratosphere (Nuclear tests) Water cycle wet Continental troposphere (re-evaporation, industry)
C.1.1 The natural radioisotopes 10Be and 210Pb 2/3 of the total 10Be production occur in the stratosphere (Lal and Peters 1967), with a spatial maximum at the magnetic poles of the earth. With a residence time in the stratosphere of about 1.3 yr (Raisbeck et al. 1981) 10Be enters the troposphere in the course of the stratosphere - troposphere exchange. The exchange is probably strongest in the middle latitudes and is expected to exhibit temporal maximum during the growing season. According to the mesurements of Luder (1989) 10Be concentrations in precipitation at Jungfraujoch (Bernese Alps, 3580 m asl.) exhibit a slight seasonal maximum in summer. VI
Stratosphere 137Cs 10Be 3H
Stratosphere - Troposphere exchange
Free troposphere Colle Gnifetti
Convective mixing
Boundary layer 3H 210Pb
Figure C.1: Sources of radionuclides and their transport to high elevation alpine sites.
The noble gas 222Rn as a decay product of 238U is exhaled from the ground. Its decay in the atmosphere produces via short lived nuclei 210Pb. Thus the ice free continents are the main source of atmospheric 210Pb (Armengaud and Genthon 1994). As the transport to the sampling site is controlled by the seasonally variable convective mixing, seasonal variations are expected to resemble the strong variations of the mainly terrigenic SO4. Figure C.2 shows the seasonal variations of SO4 concentrations in an ice core at CDD (Preunkert et al. 2000). In fact 210Pb snow concentration at Colle Gnifetti exhibit a summer level of 6.4 dpm/kg and a winter level of 1.77 – 3.19 dpm/kg (Völker 1996). Virtually right after their production both 10Be and 210Pb are attached to atmospheric sub- micron aerosols (Thurekian et al. 1989). Together with the aerosols 10Be and 210Pb are removed from the atmosphere by wet and dry deposition. VII
1000 ] b p p [
4 O S
100 1 2 3 4 5 6 7 8 9 10 11 12 13 portion of the year
Figure C.2: Seasonal variations of SO4 in an ice core from CDD (Preunkert et al. 2000)
C.1.2 The artificial radioisotopes 137Cs and Tritium The major part of their cumulative deposition during the last decades was produced by nuclear weapons between 1945 and 1980, while the maximum of the deposition rate in the northern hemisphere occurred in 1963. 137Cs was injected into the stratosphere mainly as a fission product of 238U. Mid-latitude tropospheric air concentrations of 137Cs show a seasonal maximum in late spring. Like 10Be and 210Pb, 137Cs is bound to aerosols, however its smaller atmospheric residence time compared to 10Be suggests that it is partly attached to larger dust particles formed during the nuclear tests (Luder 1986).
Tritium entered the stratosphere mainly as a product of fusion bombs. Like for 137Cs, the maximum in the north hemispheric precipitation occured in 1963. However, as tritium is a part of hydrological cycle continental re-evaporation became a source of tritium in the following years. Moreover, industrial sources are found within the Alpine region. Thus it is expected that the main source has shifted from the stratosphere in the years of maximum nuclear weapon test activity to the continental troposphere. Consequently the seasonal variations may have changed from resembling 10Be to resembling 210Pb. In fact the annual tritium maximum observed in an ice core from Fiescherhorn (Bernese Alps, 4000 m a.s.l.) is almost simultaneous with the maximum in stable isotopes during the 60th ‘s (Schiwikowski et al. 1999) but it occurs before the seasonal maximum of stable isotopes for the core sections 1977-1978 (Schotterer et al. 1998). Moreover this core indicates that a considerable portion of the total inventory was deposited in 1963 with a unusual seasonality. VIII
C.2 Methods
From the cores KCH and KCS (CG) and the core C10 (CDD) bulk samples covering definite time periods were prepared to determine the inventories of 10Be, 210Pb, 3H and 137Cs. The time ranges covered by the bulk samples were chosen to be: 1. 1950 to arounfd 1984 in order to include the main surface tests of nuclear weapons but to exclude the 3H and 137Cs input of the Chernobyl accident 1986. The later was excluded since as a single tropospheric event its impact on the total inventory would be strongly affected by air mass transport and depositional noise. 2. 1984 to uppermost core section.
From all core sections in the time range a slice of equal cross section was cut parallel to the core axis. Since the 137Cs and 3H concentrations exhibit variations of several orders of magnitude in the considered time range, the inventories are very sensitive with respect to variations in the cross section of the slices. In order to minimize the effects of variable cross sections, for each core section the target mass of the slice was calculated proportional to the total mass of the core section. The actual weight of the slice was adjusted to the target value by first cutting a slightly thicker slice and then scraping off and weighting it if needed repeatedly. Figure C.3 shows an overview of the radionuclide sample preparation procedure. The bulk sample of total mass 1-2 kg was melted in a PE bag and 9 ml of water were set aside for measuring tritium by gas proportional counting. Then to eliminate wall effects the melted sample was acidified to pH ~1.5 by adding 1-2 ml of concentrated HCl. As a yield tracer for 210Po and as carrier for 9Be 0.08-0.25 ml of 12 dpm/kg 209Po and 0.2 -1 ml of 1000 ppm 9Be were added to the sample. After evaporation to ≈ 40 g the 137Cs content was measured by γ- spectroscopy. For measuring the 210Pb concentration, its daughter nuclide in decay equilibrium 210Po is collected by self-deposition on a silver disc dipped into the solution. 210Po is then measured by α-spectroscopy. To determine the 10Be content the isobar 10B has to be removed since it would make the AMS measurement impossible. For that pH ≈ 8.5 is adjusted by adding NH3. This causes Be(OH)2 to precipitate, whereas B is still in solution. The sample is centrifuged and the water removed.
At a temperature of 1000°C Be(OH)2 is then oxidised to BeO and pressed onto a Cu-target for the AMS measurement of the 10Be/9Be ratio. A more detailed description of the radionuclide sample preparation for 10Be and 210Pb analysis can be found at Stanzick (2001) and Stanzick (1996). IX
Bulk sample Aliquot for Melting tritium analysis
Add HCl add Spike Evaporation
γ-spectroscopy 1 3 7 of Cs
Deposition 2 1 0 2 1 0 of Po α-spectroscopy of Po on a silver disc
Precipitation 1 0 AMS target for Be of Be(OH) 2
Figure C.3: Radioisotope sample preparation procedure
C.3 Results
All inventories are shown in Table C.2. Note that the surface accumulation rate is 0.24 m w.e. for KCH, 0.52 m w.e. for KCS and 2.5 m w.e. for C10.
Table C.2: Decadal means of 210Pb and 10Be in deep cores from CG and CDD. All activities refer to 01.01.2000. (1): below detection limit. (2): not yet measured.
Core Time slice 10Be [107 at/kg] 210Pb [dpm/kg] 137Cs [fCi/g] Tritium [TU]
KCH 1983 – 1993 1.91 ± 0.1 6.4 ± 0.3 (1) 8.2 ± 1.5
KCH 1950 – 1983 2.04 ± 0.1 5.0 ± 0.3 2.2 ± 0.2 59 ± 2.2
KCS 1983 – 1992 1.69 ± 0.1 7.9 ± 0.4 7.3 ± 0.8 8.3 ± 1.6
KCS 1951 - 1983 1.55 ± 0.1 5.7 ± 0.3 1.8 ± 0.2 41 ± 2
C 10 1985 - 1991 1.79 ± 0.1 4.0 ± 0.3 (1) 5.0 ± 1.4
C 10 1965 -1985 1.63 ± 0.1 3.4 ± 0.2 0.65 ± 0.06 (2)
C 10 1945 - 1965 1.86 ± 0.1 11.4 ± 0.6 3.7 ± 0.5 78 ± 3 X
10Be As expected no temporal trend is observed in 10Be. 10Be concentrations of KCH are around 30% higher than for KCS and C10. Because KCH has the lowest accumulation rate and as the seasonal variations in 10Be concentrations are expected to be rather small, the proportionally higher contribution of dry deposition probably causes the enhanced level for KCH.
210Pb The concentrations of KCS 1983 – 1992 and C10 1945 – 1965 are above the typical summer level at CG. These samples may have been subject to contamination. All other concentrations from CG concurrently resemble the typical summer level there while the concentrations for C10 are lower. This is partly caused by the higher percentage of summer snow in the CG cores, but also dry deposition may contribute to the difference between CG and the high accumulation core C10. The later reason is supported by the finding that cores of higher accumulation rate tend to show a lower mean 210Pb (Völker 1996). As the source strength of 210Pb is expected to be constant, the concurrent increase of 210Pb in the younger time slice of the cores KCH and C10 may suggest that the tropospheric vertical mixing intensity has increased. This is in agreement with the finding that the frequency of convective weather types has increased in the course of the warming in the second half of the 20th century (Stefanicki et al. 1998).
137Cs The results of the 137Cs inventories do not fit with the expectations. The values below detection limit suggest a yield problem.
Tritium The similar concentrations of comparable time slices for KCH and KCS suggest that the recorded season of these two cores is not much different (compare Figure 6.5). The higher values for KCH in the time slice 1950 – 1983 are hard to interpret since the source has changed in the course of time. XI
Annex D: Glacier simulator experiments
After the suspicion arose that somehow or other gradients in ice deformation cause the migration of stable isotopes in the ice matrix, a direct proof by laboratory experiment was attempted. However, a plausible idea for the process came up much later. The potential process suggested in 8.4 can probably not be simulated with such simple laboratory experiments. Accordingly this is the report of a failed attempt. Glacier simulators were frequently used to investigate ice deformation under well defined conditions (e.g. Meyssonnier 1989). The basic idea of the experiment is to fill a glacier simulator with isotopically homogeneous polycristalline ice. The isotope profiles from intensely deformed ice are then compared with the corresponding profiles from not deformed ice.
D.1 Production of homogeneous polycristalline ice
First the production of artificial, polycristalline ice was established. For that Alpine cold firn (δD ≈ -150 ‰) was filled in a clean PE-bag. It was grinded up in single grains and well mixed by mechanical squeezing of the bag. 350 g of these ice crystals were brought in a commercial ice-cream maker with a cooling container being tempered slightly below 0°C. During continuous stirring 100 ml cooled de-ionized water (δD ≈ -60 ‰) was dripped on the firn until the ice pap appeared as homogeneous slush. This ice pap could be handily spooned in the desired vessels. For investigations of the isotopic homogeneity a 250 ml vessel was filled with the ice pap and frozen over night. The product was a optically homogeneous, dull, bubble containing ice of density 0.87 g/cm3 and a crystal size around 1 mm. From this artificial ice two sample series of mass 2 g and 0.2 g respectively were cut by saw and scalpel respectively. The standard deviations turned out to be 0.4 ‰ for the 2 g samples and 1.2 ‰ for the 0.2 g samples.
D.2 Design and operation of the glacier simulator
The apparatus for ice deformation is shown in Figure D.1. The notched inner axis is fixed. The ice is set in motion by applying torque on an outer notched ring using a weight hanging on a rope which is wrapped on the cylinder. Top and bottom of the apparatus were connected to the outer casing. XII
rotatable casing
fixed axis
ice pap rope
10 cm
Figure D.1. Cross-section of the apparatus for ice deformation
The shear stress equals to:
m ⋅ g ⋅ R τ (r) = with: 2 ⋅π ⋅ r 2 ⋅ h τ (r) = shear stress at radius r m = mass of the weight (27.5 kg) R = outer radius of the cylinder (5.5 cm) H = height of the cylinder (3.5 cm)
Thus the initial stress maximum at the tips of the notches at the axis (r = 1.15 cm) equals around 4.7 ⋅ 105 Pa. This is well above the typical basal shear stress of around 1 ⋅ 105 Pa (Paterson 1994). The glacier simulator was set up in an insulated box inside the cold room. Temperature was adjusted to –7 ± 0.5°C by a 15 W bulb inside the box connected to a dimmer. The decrease in pressure melting point caused by the mentioned stress maximum is around 0.04°C. Therefore any melting during the operation of the simulator can be excluded. The apparatus was filled with the artificial ice pap. After being frozen completely stress was applied and ice deformation, noted by the turning outer casing, started. After 5 days ice deformation was rapidly increasing and the maximum stress was diminished from 4.7 to 2.8 bar. After that the casing kept on turning at a rate of around 15°/day and after 3 weeks and having one revolution completed it was stopped (see Figure D.2). XIII
80 ] d / ° [
ω 40
0 0 5 10 15 20 25 time [d]
Figure D.2: Angular velocity of the turning glacier simulator. The low values after 3.8 days and 5.2 days were caused by mechanical problems.
D.3 Results and discussion
The filling which was subject to deformation was compared to two not deformed reverence fillings. The radial symmetric stratigraphic features caused by the deformation are listed in Table D.1.
Table D.1: Stratigraphic features caused by ice deformation
Section Radius Stratigraphy
I 0.6 – 1.1 (between notches) Clear ice without bubbles II 1.1 – 2.2 Clear ice with bubbles III 2.2 – 5.4 Dull ice without bubbles (apparently unchanged)
To investigate the crucial question, that is whether any significant effect of the ice deformation on the distribution of stable isotopes can be detected, δD values from three radial profiles from the deformed filling were compared to three profiles from the not deformed fillings. The result is presented in Figure D.3. XIV
: deformed : not deformed
5 I II III g n i s ] s i a x c ‰
[ a r
e D t δ u
o ∆
0
0 1 2 3 4 5 radius [cm]
Figure D.3: Deviation from the mean isotopic composition of the respective filling. Filled squares, circles and triangles refer to the three profiles from the deformed filling, the open symbols refer to the undeformed profiles. I, II and III refer to the sections fromTable D.1
The systematically higher δD values between the axis and r ≈ 1.5 cm occur no matter whether the ice was subject to shear or not. Therefore they can not be attributed to ice deformation. That enhanced δD values near the walls were not observed in the preliminary homogeneity tests of the ice pap. The explanation is probably that due the slight giggling of the axis when closing the glacier simulator the de-ionised water is partly separated from the firn. With the process suggested in section 8.4 in mind it is clear that the migration of isotopes can not be proofed by such a simple experiment. The main point is that the small volume fraction of cavities, i.e. the bubbles in section I and II are filled just once. What would be needed is to simulate the dynamic process of continuously generating cavities and refilling them. This seems to be pretty difficult in a laboratory experiment. XV
Acknowledgements
Prof. Dr. Ulrich Platt and Prof. Dr. Konrad Mauersberger kindly appraised this thesis. Dr. Dietmar Wagenbach is highly appreciated for his extraordinary systematic overview and numerous helpful comments on this work. My PhD companions Michael Huke, Urs Ruth and Andreas Stanzick provided a pleasant atmosphere in the Heidelberg glaciology group and were very helpful both for laboratory and field work. Marc Armbruster did a good job in establishing the chemical profiles and in dating. The stable isotope measurements benefited a lot from the routine of Christel Facklam. Field activities were very much supported by Stephan Suter (VAW), Stefan Gruber, Martin Hoelzle and several other co-workers. Martin Lüthi (also VAW) has substantially contributed to the survey and the interpretation of ice core records by the 3D flow model for Colle Gnifetti. Hermann Boesch has patiently evaluated survey data. Radar profiles were measured and evaluated by Uwe Nixdorf and Olaf Eisen (AWI Bremerhaven). I am also very grateful to the Swiss mountain guides Gabriel Willisch, Bruno Wyrsch and in particular to Mr. Fredi Biner. Probably I did not have to watch a tragedy at Colle Gnifetti due to his competence.